lighthouse/beacon_node/beacon_chain/tests/store_tests.rs

2744 lines
91 KiB
Rust
Raw Normal View History

#![cfg(not(debug_assertions))]
Add attestation gossip pre-verification (#983) * Add PH & MS slot clock changes * Account for genesis time * Add progress on duties refactor * Add simple is_aggregator bool to val subscription * Start work on attestation_verification.rs * Add progress on ObservedAttestations * Progress with ObservedAttestations * Fix tests * Add observed attestations to the beacon chain * Add attestation observation to processing code * Add progress on attestation verification * Add first draft of ObservedAttesters * Add more tests * Add observed attesters to beacon chain * Add observers to attestation processing * Add more attestation verification * Create ObservedAggregators map * Remove commented-out code * Add observed aggregators into chain * Add progress * Finish adding features to attestation verification * Ensure beacon chain compiles * Link attn verification into chain * Integrate new attn verification in chain * Remove old attestation processing code * Start trying to fix beacon_chain tests * Split adding into pools into two functions * Add aggregation to harness * Get test harness working again * Adjust the number of aggregators for test harness * Fix edge-case in harness * Integrate new attn processing in network * Fix compile bug in validator_client * Update validator API endpoints * Fix aggreagation in test harness * Fix enum thing * Fix attestation observation bug: * Patch failing API tests * Start adding comments to attestation verification * Remove unused attestation field * Unify "is block known" logic * Update comments * Supress fork choice errors for network processing * Add todos * Tidy * Add gossip attn tests * Disallow test harness to produce old attns * Comment out in-progress tests * Partially address pruning tests * Fix failing store test * Add aggregate tests * Add comments about which spec conditions we check * Dont re-aggregate * Split apart test harness attn production * Fix compile error in network * Make progress on commented-out test * Fix skipping attestation test * Add fork choice verification tests * Tidy attn tests, remove dead code * Remove some accidentally added code * Fix clippy lint * Rename test file * Add block tests, add cheap block proposer check * Rename block testing file * Add observed_block_producers * Tidy * Switch around block signature verification * Finish block testing * Remove gossip from signature tests * First pass of self review * Fix deviation in spec * Update test spec tags * Start moving over to hashset * Finish moving observed attesters to hashmap * Move aggregation pool over to hashmap * Make fc attn borrow again * Fix rest_api compile error * Fix missing comments * Fix monster test * Uncomment increasing slots test * Address remaining comments * Remove unsafe, use cfg test * Remove cfg test flag * Fix dodgy comment * Ignore aggregates that are already known. * Unify aggregator modulo logic * Fix typo in logs * Refactor validator subscription logic * Avoid reproducing selection proof * Skip HTTP call if no subscriptions * Rename DutyAndState -> DutyAndProof * Tidy logs * Print root as dbg * Fix compile errors in tests * Fix compile error in test
2020-05-06 11:42:56 +00:00
use beacon_chain::attestation_verification::Error as AttnError;
use beacon_chain::builder::BeaconChainBuilder;
use beacon_chain::test_utils::{
Add test_logger as feature to logging (#2586) ## Issue Addressed Fix #2585 ## Proposed Changes Provide a canonical version of test_logger that can be used throughout lighthouse. ## Additional Info This allows tests to conditionally emit logging data by adding test_logger as the default logger. And then when executing `cargo test --features logging/test_logger` log output will be visible: wink@3900x:~/lighthouse/common/logging/tests/test-feature-test_logger (Add-test_logger-as-feature-to-logging) $ cargo test --features logging/test_logger Finished test [unoptimized + debuginfo] target(s) in 0.02s Running unittests (target/debug/deps/test_logger-e20115db6a5e3714) running 1 test Sep 10 12:53:45.212 INFO hi, module: test_logger:8 test tests::test_fn_with_logging ... ok test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s Doc-tests test-logger running 0 tests test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s Or, in normal scenarios where logging isn't needed, executing `cargo test` the log output will not be visible: wink@3900x:~/lighthouse/common/logging/tests/test-feature-test_logger (Add-test_logger-as-feature-to-logging) $ cargo test Finished test [unoptimized + debuginfo] target(s) in 0.02s Running unittests (target/debug/deps/test_logger-02e02f8d41e8cf8a) running 1 test test tests::test_fn_with_logging ... ok test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s Doc-tests test-logger running 0 tests test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
2021-10-06 00:46:07 +00:00
test_spec, AttestationStrategy, BeaconChainHarness, BlockStrategy, DiskHarnessType,
};
use beacon_chain::{
historical_blocks::HistoricalBlockError, migrate::MigratorConfig, BeaconChain,
BeaconChainError, BeaconChainTypes, BeaconSnapshot, ChainConfig, NotifyExecutionLayer,
ServerSentEventHandler, WhenSlotSkipped,
};
use fork_choice::CountUnrealized;
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
use lazy_static::lazy_static;
Add test_logger as feature to logging (#2586) ## Issue Addressed Fix #2585 ## Proposed Changes Provide a canonical version of test_logger that can be used throughout lighthouse. ## Additional Info This allows tests to conditionally emit logging data by adding test_logger as the default logger. And then when executing `cargo test --features logging/test_logger` log output will be visible: wink@3900x:~/lighthouse/common/logging/tests/test-feature-test_logger (Add-test_logger-as-feature-to-logging) $ cargo test --features logging/test_logger Finished test [unoptimized + debuginfo] target(s) in 0.02s Running unittests (target/debug/deps/test_logger-e20115db6a5e3714) running 1 test Sep 10 12:53:45.212 INFO hi, module: test_logger:8 test tests::test_fn_with_logging ... ok test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s Doc-tests test-logger running 0 tests test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s Or, in normal scenarios where logging isn't needed, executing `cargo test` the log output will not be visible: wink@3900x:~/lighthouse/common/logging/tests/test-feature-test_logger (Add-test_logger-as-feature-to-logging) $ cargo test Finished test [unoptimized + debuginfo] target(s) in 0.02s Running unittests (target/debug/deps/test_logger-02e02f8d41e8cf8a) running 1 test test tests::test_fn_with_logging ... ok test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s Doc-tests test-logger running 0 tests test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
2021-10-06 00:46:07 +00:00
use logging::test_logger;
use maplit::hashset;
use rand::Rng;
use state_processing::BlockReplayer;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
use std::collections::HashMap;
use std::collections::HashSet;
use std::convert::TryInto;
use std::sync::Arc;
use std::time::Duration;
use store::{
iter::{BlockRootsIterator, StateRootsIterator},
HotColdDB, LevelDB, StoreConfig,
};
use tempfile::{tempdir, TempDir};
use tokio::time::sleep;
use tree_hash::TreeHash;
use types::test_utils::{SeedableRng, XorShiftRng};
use types::*;
// Should ideally be divisible by 3.
pub const LOW_VALIDATOR_COUNT: usize = 24;
pub const HIGH_VALIDATOR_COUNT: usize = 64;
lazy_static! {
/// A cached set of keys.
static ref KEYPAIRS: Vec<Keypair> = types::test_utils::generate_deterministic_keypairs(HIGH_VALIDATOR_COUNT);
}
type E = MinimalEthSpec;
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
type TestHarness = BeaconChainHarness<DiskHarnessType<E>>;
fn get_store(db_path: &TempDir) -> Arc<HotColdDB<E, LevelDB<E>, LevelDB<E>>> {
get_store_with_spec(db_path, test_spec::<E>())
}
fn get_store_with_spec(
db_path: &TempDir,
spec: ChainSpec,
) -> Arc<HotColdDB<E, LevelDB<E>, LevelDB<E>>> {
let hot_path = db_path.path().join("hot_db");
let cold_path = db_path.path().join("cold_db");
let config = StoreConfig::default();
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
let log = test_logger();
HotColdDB::open(&hot_path, &cold_path, |_, _, _| Ok(()), config, spec, log)
.expect("disk store should initialize")
}
fn get_harness(
store: Arc<HotColdDB<E, LevelDB<E>, LevelDB<E>>>,
validator_count: usize,
) -> TestHarness {
let harness = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.keypairs(KEYPAIRS[0..validator_count].to_vec())
.fresh_disk_store(store)
.mock_execution_layer()
.build();
harness.advance_slot();
harness
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn full_participation_no_skips() {
let num_blocks_produced = E::slots_per_epoch() * 5;
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
let harness = get_harness(store.clone(), LOW_VALIDATOR_COUNT);
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
num_blocks_produced as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
check_finalization(&harness, num_blocks_produced);
check_split_slot(&harness, store);
check_chain_dump(&harness, num_blocks_produced + 1);
check_iterators(&harness);
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn randomised_skips() {
let num_slots = E::slots_per_epoch() * 5;
let mut num_blocks_produced = 0;
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
let harness = get_harness(store.clone(), LOW_VALIDATOR_COUNT);
let rng = &mut XorShiftRng::from_seed([42; 16]);
let mut head_slot = 0;
for slot in 1..=num_slots {
if rng.gen_bool(0.8) {
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
1,
BlockStrategy::ForkCanonicalChainAt {
previous_slot: Slot::new(head_slot),
first_slot: Slot::new(slot),
},
AttestationStrategy::AllValidators,
)
.await;
harness.advance_slot();
num_blocks_produced += 1;
head_slot = slot;
} else {
harness.advance_slot();
}
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let state = &harness.chain.head_snapshot().beacon_state;
assert_eq!(
state.slot(),
num_slots,
"head should be at the current slot"
);
check_split_slot(&harness, store.clone());
check_chain_dump(&harness, num_blocks_produced + 1);
check_iterators(&harness);
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn long_skip() {
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
let harness = get_harness(store.clone(), LOW_VALIDATOR_COUNT);
// Number of blocks to create in the first run, intentionally not falling on an epoch
// boundary in order to check that the DB hot -> cold migration is capable of reaching
// back across the skip distance, and correctly migrating those extra non-finalized states.
let initial_blocks = E::slots_per_epoch() * 5 + E::slots_per_epoch() / 2;
let skip_slots = E::slots_per_historical_root() as u64 * 8;
// Create the minimum ~2.5 epochs of extra blocks required to re-finalize the chain.
// Having this set lower ensures that we start justifying and finalizing quickly after a skip.
let final_blocks = 2 * E::slots_per_epoch() + E::slots_per_epoch() / 2;
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
initial_blocks as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
check_finalization(&harness, initial_blocks);
// 2. Skip slots
for _ in 0..skip_slots {
harness.advance_slot();
}
// 3. Produce more blocks, establish a new finalized epoch
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
final_blocks as usize,
BlockStrategy::ForkCanonicalChainAt {
previous_slot: Slot::new(initial_blocks),
first_slot: Slot::new(initial_blocks + skip_slots as u64 + 1),
},
AttestationStrategy::AllValidators,
)
.await;
check_finalization(&harness, initial_blocks + skip_slots + final_blocks);
check_split_slot(&harness, store);
check_chain_dump(&harness, initial_blocks + final_blocks + 1);
check_iterators(&harness);
}
/// Go forward to the point where the genesis randao value is no longer part of the vector.
///
/// This implicitly checks that:
/// 1. The chunked vector scheme doesn't attempt to store an incorrect genesis value
/// 2. We correctly load the genesis value for all required slots
/// NOTE: this test takes about a minute to run
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn randao_genesis_storage() {
let validator_count = 8;
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
let harness = get_harness(store.clone(), validator_count);
let num_slots = E::slots_per_epoch() * (E::epochs_per_historical_vector() - 1) as u64;
// Check we have a non-trivial genesis value
let genesis_value = *harness
.chain
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
.head_snapshot()
.beacon_state
.get_randao_mix(Epoch::new(0))
.expect("randao mix ok");
assert!(!genesis_value.is_zero());
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
num_slots as usize - 1,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
// Check that genesis value is still present
assert!(harness
.chain
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
.head_snapshot()
.beacon_state
.randao_mixes()
.iter()
.find(|x| **x == genesis_value)
.is_some());
// Then upon adding one more block, it isn't
harness.advance_slot();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
1,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
assert!(harness
.chain
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
.head_snapshot()
.beacon_state
.randao_mixes()
.iter()
.find(|x| **x == genesis_value)
.is_none());
check_finalization(&harness, num_slots);
check_split_slot(&harness, store);
check_chain_dump(&harness, num_slots + 1);
check_iterators(&harness);
}
// Check that closing and reopening a freezer DB restores the split slot to its correct value.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn split_slot_restore() {
let db_path = tempdir().unwrap();
let split_slot = {
let store = get_store(&db_path);
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
let harness = get_harness(store.clone(), LOW_VALIDATOR_COUNT);
let num_blocks = 4 * E::slots_per_epoch();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
num_blocks as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
store.get_split_slot()
};
assert_ne!(split_slot, Slot::new(0));
// Re-open the store
let store = get_store(&db_path);
assert_eq!(store.get_split_slot(), split_slot);
}
// Check attestation processing and `load_epoch_boundary_state` in the presence of a split DB.
// This is a bit of a monster test in that it tests lots of different things, but until they're
// tested elsewhere, this is as good a place as any.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn epoch_boundary_state_attestation_processing() {
let num_blocks_produced = E::slots_per_epoch() * 5;
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
let harness = get_harness(store.clone(), LOW_VALIDATOR_COUNT);
let late_validators = vec![0, 1];
let timely_validators = (2..LOW_VALIDATOR_COUNT).collect::<Vec<_>>();
let mut late_attestations = vec![];
for _ in 0..num_blocks_produced {
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
1,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::SomeValidators(timely_validators.clone()),
)
.await;
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let head = harness.chain.head_snapshot();
Add attestation gossip pre-verification (#983) * Add PH & MS slot clock changes * Account for genesis time * Add progress on duties refactor * Add simple is_aggregator bool to val subscription * Start work on attestation_verification.rs * Add progress on ObservedAttestations * Progress with ObservedAttestations * Fix tests * Add observed attestations to the beacon chain * Add attestation observation to processing code * Add progress on attestation verification * Add first draft of ObservedAttesters * Add more tests * Add observed attesters to beacon chain * Add observers to attestation processing * Add more attestation verification * Create ObservedAggregators map * Remove commented-out code * Add observed aggregators into chain * Add progress * Finish adding features to attestation verification * Ensure beacon chain compiles * Link attn verification into chain * Integrate new attn verification in chain * Remove old attestation processing code * Start trying to fix beacon_chain tests * Split adding into pools into two functions * Add aggregation to harness * Get test harness working again * Adjust the number of aggregators for test harness * Fix edge-case in harness * Integrate new attn processing in network * Fix compile bug in validator_client * Update validator API endpoints * Fix aggreagation in test harness * Fix enum thing * Fix attestation observation bug: * Patch failing API tests * Start adding comments to attestation verification * Remove unused attestation field * Unify "is block known" logic * Update comments * Supress fork choice errors for network processing * Add todos * Tidy * Add gossip attn tests * Disallow test harness to produce old attns * Comment out in-progress tests * Partially address pruning tests * Fix failing store test * Add aggregate tests * Add comments about which spec conditions we check * Dont re-aggregate * Split apart test harness attn production * Fix compile error in network * Make progress on commented-out test * Fix skipping attestation test * Add fork choice verification tests * Tidy attn tests, remove dead code * Remove some accidentally added code * Fix clippy lint * Rename test file * Add block tests, add cheap block proposer check * Rename block testing file * Add observed_block_producers * Tidy * Switch around block signature verification * Finish block testing * Remove gossip from signature tests * First pass of self review * Fix deviation in spec * Update test spec tags * Start moving over to hashset * Finish moving observed attesters to hashmap * Move aggregation pool over to hashmap * Make fc attn borrow again * Fix rest_api compile error * Fix missing comments * Fix monster test * Uncomment increasing slots test * Address remaining comments * Remove unsafe, use cfg test * Remove cfg test flag * Fix dodgy comment * Ignore aggregates that are already known. * Unify aggregator modulo logic * Fix typo in logs * Refactor validator subscription logic * Avoid reproducing selection proof * Skip HTTP call if no subscriptions * Rename DutyAndState -> DutyAndProof * Tidy logs * Print root as dbg * Fix compile errors in tests * Fix compile error in test
2020-05-06 11:42:56 +00:00
late_attestations.extend(harness.get_unaggregated_attestations(
&AttestationStrategy::SomeValidators(late_validators.clone()),
&head.beacon_state,
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
head.beacon_state_root(),
head.beacon_block_root,
Update to Spec v0.10 (#817) * Start updating types * WIP * Signature hacking * Existing EF tests passing with fake_crypto * Updates * Delete outdated API spec * The refactor continues * It compiles * WIP test fixes * All release tests passing bar genesis state parsing * Update and test YamlConfig * Update to spec v0.10 compatible BLS * Updates to BLS EF tests * Add EF test for AggregateVerify And delete unused hash2curve tests for uncompressed points * Update EF tests to v0.10.1 * Use optional block root correctly in block proc * Use genesis fork in deposit domain. All tests pass * Cargo fmt * Fast aggregate verify test * Update REST API docs * Cargo fmt * Fix unused import * Bump spec tags to v0.10.1 * Add `seconds_per_eth1_block` to chainspec * Update to timestamp based eth1 voting scheme * Return None from `get_votes_to_consider` if block cache is empty * Handle overflows in `is_candidate_block` * Revert to failing tests * Fix eth1 data sets test * Choose default vote according to spec * Fix collect_valid_votes tests * Fix `get_votes_to_consider` to choose all eligible blocks * Uncomment winning_vote tests * Add comments; remove unused code * Reduce seconds_per_eth1_block for simulation * Addressed review comments * Add test for default vote case * Fix logs * Remove unused functions * Meter default eth1 votes * Fix comments * Address review comments; remove unused dependency * Disable/delete two outdated tests * Bump eth1 default vote warn to error * Delete outdated eth1 test Co-authored-by: Pawan Dhananjay <pawandhananjay@gmail.com>
2020-02-10 23:19:36 +00:00
head.beacon_block.slot(),
));
harness.advance_slot();
}
check_finalization(&harness, num_blocks_produced);
check_split_slot(&harness, store.clone());
check_chain_dump(&harness, num_blocks_produced + 1);
check_iterators(&harness);
let mut checked_pre_fin = false;
for (attestation, subnet_id) in late_attestations.into_iter().flatten() {
// load_epoch_boundary_state is idempotent!
let block_root = attestation.data.beacon_block_root;
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
let block = store
.get_blinded_block(&block_root)
.unwrap()
.expect("block exists");
let epoch_boundary_state = store
Update to Spec v0.10 (#817) * Start updating types * WIP * Signature hacking * Existing EF tests passing with fake_crypto * Updates * Delete outdated API spec * The refactor continues * It compiles * WIP test fixes * All release tests passing bar genesis state parsing * Update and test YamlConfig * Update to spec v0.10 compatible BLS * Updates to BLS EF tests * Add EF test for AggregateVerify And delete unused hash2curve tests for uncompressed points * Update EF tests to v0.10.1 * Use optional block root correctly in block proc * Use genesis fork in deposit domain. All tests pass * Cargo fmt * Fast aggregate verify test * Update REST API docs * Cargo fmt * Fix unused import * Bump spec tags to v0.10.1 * Add `seconds_per_eth1_block` to chainspec * Update to timestamp based eth1 voting scheme * Return None from `get_votes_to_consider` if block cache is empty * Handle overflows in `is_candidate_block` * Revert to failing tests * Fix eth1 data sets test * Choose default vote according to spec * Fix collect_valid_votes tests * Fix `get_votes_to_consider` to choose all eligible blocks * Uncomment winning_vote tests * Add comments; remove unused code * Reduce seconds_per_eth1_block for simulation * Addressed review comments * Add test for default vote case * Fix logs * Remove unused functions * Meter default eth1 votes * Fix comments * Address review comments; remove unused dependency * Disable/delete two outdated tests * Bump eth1 default vote warn to error * Delete outdated eth1 test Co-authored-by: Pawan Dhananjay <pawandhananjay@gmail.com>
2020-02-10 23:19:36 +00:00
.load_epoch_boundary_state(&block.state_root())
.expect("no error")
.expect("epoch boundary state exists");
let ebs_of_ebs = store
.load_epoch_boundary_state(&epoch_boundary_state.canonical_root())
.expect("no error")
.expect("ebs of ebs exists");
assert_eq!(epoch_boundary_state, ebs_of_ebs);
// If the attestation is pre-finalization it should be rejected.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let finalized_epoch = harness.finalized_checkpoint().epoch;
Add attestation gossip pre-verification (#983) * Add PH & MS slot clock changes * Account for genesis time * Add progress on duties refactor * Add simple is_aggregator bool to val subscription * Start work on attestation_verification.rs * Add progress on ObservedAttestations * Progress with ObservedAttestations * Fix tests * Add observed attestations to the beacon chain * Add attestation observation to processing code * Add progress on attestation verification * Add first draft of ObservedAttesters * Add more tests * Add observed attesters to beacon chain * Add observers to attestation processing * Add more attestation verification * Create ObservedAggregators map * Remove commented-out code * Add observed aggregators into chain * Add progress * Finish adding features to attestation verification * Ensure beacon chain compiles * Link attn verification into chain * Integrate new attn verification in chain * Remove old attestation processing code * Start trying to fix beacon_chain tests * Split adding into pools into two functions * Add aggregation to harness * Get test harness working again * Adjust the number of aggregators for test harness * Fix edge-case in harness * Integrate new attn processing in network * Fix compile bug in validator_client * Update validator API endpoints * Fix aggreagation in test harness * Fix enum thing * Fix attestation observation bug: * Patch failing API tests * Start adding comments to attestation verification * Remove unused attestation field * Unify "is block known" logic * Update comments * Supress fork choice errors for network processing * Add todos * Tidy * Add gossip attn tests * Disallow test harness to produce old attns * Comment out in-progress tests * Partially address pruning tests * Fix failing store test * Add aggregate tests * Add comments about which spec conditions we check * Dont re-aggregate * Split apart test harness attn production * Fix compile error in network * Make progress on commented-out test * Fix skipping attestation test * Add fork choice verification tests * Tidy attn tests, remove dead code * Remove some accidentally added code * Fix clippy lint * Rename test file * Add block tests, add cheap block proposer check * Rename block testing file * Add observed_block_producers * Tidy * Switch around block signature verification * Finish block testing * Remove gossip from signature tests * First pass of self review * Fix deviation in spec * Update test spec tags * Start moving over to hashset * Finish moving observed attesters to hashmap * Move aggregation pool over to hashmap * Make fc attn borrow again * Fix rest_api compile error * Fix missing comments * Fix monster test * Uncomment increasing slots test * Address remaining comments * Remove unsafe, use cfg test * Remove cfg test flag * Fix dodgy comment * Ignore aggregates that are already known. * Unify aggregator modulo logic * Fix typo in logs * Refactor validator subscription logic * Avoid reproducing selection proof * Skip HTTP call if no subscriptions * Rename DutyAndState -> DutyAndProof * Tidy logs * Print root as dbg * Fix compile errors in tests * Fix compile error in test
2020-05-06 11:42:56 +00:00
let res = harness
.chain
Batch BLS verification for attestations (#2399) ## Issue Addressed NA ## Proposed Changes Adds the ability to verify batches of aggregated/unaggregated attestations from the network. When the `BeaconProcessor` finds there are messages in the aggregated or unaggregated attestation queues, it will first check the length of the queue: - `== 1` verify the attestation individually. - `>= 2` take up to 64 of those attestations and verify them in a batch. Notably, we only perform batch verification if the queue has a backlog. We don't apply any artificial delays to attestations to try and force them into batches. ### Batching Details To assist with implementing batches we modify `beacon_chain::attestation_verification` to have two distinct categories for attestations: - *Indexed* attestations: those which have passed initial validation and were valid enough for us to derive an `IndexedAttestation`. - *Verified* attestations: those attestations which were indexed *and also* passed signature verification. These are well-formed, interesting messages which were signed by validators. The batching functions accept `n` attestations and then return `n` attestation verification `Result`s, where those `Result`s can be any combination of `Ok` or `Err`. In other words, we attempt to verify as many attestations as possible and return specific per-attestation results so peer scores can be updated, if required. When we batch verify attestations, we first try to map all those attestations to *indexed* attestations. If any of those attestations were able to be indexed, we then perform batch BLS verification on those indexed attestations. If the batch verification succeeds, we convert them into *verified* attestations, disabling individual signature checking. If the batch fails, we convert to verified attestations with individual signature checking enabled. Ultimately, we optimistically try to do a batch verification of attestation signatures and fall-back to individual verification if it fails. This opens an attach vector for "poisoning" the attestations and causing us to waste a batch verification. I argue that peer scoring should do a good-enough job of defending against this and the typical-case gains massively outweigh the worst-case losses. ## Additional Info Before this PR, attestation verification took the attestations by value (instead of by reference). It turns out that this was unnecessary and, in my opinion, resulted in some undesirable ergonomics (e.g., we had to pass the attestation back in the `Err` variant to avoid clones). In this PR I've modified attestation verification so that it now takes a reference. I refactored the `beacon_chain/tests/attestation_verification.rs` tests so they use a builder-esque "tester" struct instead of a weird macro. It made it easier for me to test individual/batch with the same set of tests and I think it was a nice tidy-up. Notably, I did this last to try and make sure my new refactors to *actual* production code would pass under the existing test suite.
2021-09-22 08:49:41 +00:00
.verify_unaggregated_attestation_for_gossip(&attestation, Some(subnet_id));
Optimize attestation processing (#841) * Start updating types * WIP * Signature hacking * Existing EF tests passing with fake_crypto * Updates * Delete outdated API spec * The refactor continues * It compiles * WIP test fixes * All release tests passing bar genesis state parsing * Update and test YamlConfig * Update to spec v0.10 compatible BLS * Updates to BLS EF tests * Add EF test for AggregateVerify And delete unused hash2curve tests for uncompressed points * Update EF tests to v0.10.1 * Use optional block root correctly in block proc * Use genesis fork in deposit domain. All tests pass * Cargo fmt * Fast aggregate verify test * Update REST API docs * Cargo fmt * Fix unused import * Bump spec tags to v0.10.1 * Add `seconds_per_eth1_block` to chainspec * Update to timestamp based eth1 voting scheme * Return None from `get_votes_to_consider` if block cache is empty * Handle overflows in `is_candidate_block` * Revert to failing tests * Fix eth1 data sets test * Choose default vote according to spec * Fix collect_valid_votes tests * Fix `get_votes_to_consider` to choose all eligible blocks * Uncomment winning_vote tests * Add comments; remove unused code * Reduce seconds_per_eth1_block for simulation * Addressed review comments * Add test for default vote case * Fix logs * Remove unused functions * Meter default eth1 votes * Fix comments * Address review comments; remove unused dependency * Add first attempt at attestation proc. re-write * Add version 2 of attestation processing * Minor fixes * Add validator pubkey cache * Make get_indexed_attestation take a committee * Link signature processing into new attn verification * First working version * Ensure pubkey cache is updated * Add more metrics, slight optimizations * Clone committee cache during attestation processing * Update shuffling cache during block processing * Remove old commented-out code * Fix shuffling cache insert bug * Used indexed attestation in fork choice * Restructure attn processing, add metrics * Add more detailed metrics * Tidy, fix failing tests * Fix failing tests, tidy * Disable/delete two outdated tests * Tidy * Add pubkey cache persistence file * Add more comments * Integrate persistence file into builder * Add pubkey cache tests * Add data_dir to beacon chain builder * Remove Option in pubkey cache persistence file * Ensure consistency between datadir/data_dir * Fix failing network test * Tidy * Fix todos * Add attestation processing tests * Add another test * Only run attestation tests in release * Make attestation tests MainnetEthSpec * Address Michael's comments * Remove redundant check * Fix warning * Fix failing test Co-authored-by: Michael Sproul <micsproul@gmail.com> Co-authored-by: Pawan Dhananjay <pawandhananjay@gmail.com>
2020-03-05 06:19:35 +00:00
Add attestation gossip pre-verification (#983) * Add PH & MS slot clock changes * Account for genesis time * Add progress on duties refactor * Add simple is_aggregator bool to val subscription * Start work on attestation_verification.rs * Add progress on ObservedAttestations * Progress with ObservedAttestations * Fix tests * Add observed attestations to the beacon chain * Add attestation observation to processing code * Add progress on attestation verification * Add first draft of ObservedAttesters * Add more tests * Add observed attesters to beacon chain * Add observers to attestation processing * Add more attestation verification * Create ObservedAggregators map * Remove commented-out code * Add observed aggregators into chain * Add progress * Finish adding features to attestation verification * Ensure beacon chain compiles * Link attn verification into chain * Integrate new attn verification in chain * Remove old attestation processing code * Start trying to fix beacon_chain tests * Split adding into pools into two functions * Add aggregation to harness * Get test harness working again * Adjust the number of aggregators for test harness * Fix edge-case in harness * Integrate new attn processing in network * Fix compile bug in validator_client * Update validator API endpoints * Fix aggreagation in test harness * Fix enum thing * Fix attestation observation bug: * Patch failing API tests * Start adding comments to attestation verification * Remove unused attestation field * Unify "is block known" logic * Update comments * Supress fork choice errors for network processing * Add todos * Tidy * Add gossip attn tests * Disallow test harness to produce old attns * Comment out in-progress tests * Partially address pruning tests * Fix failing store test * Add aggregate tests * Add comments about which spec conditions we check * Dont re-aggregate * Split apart test harness attn production * Fix compile error in network * Make progress on commented-out test * Fix skipping attestation test * Add fork choice verification tests * Tidy attn tests, remove dead code * Remove some accidentally added code * Fix clippy lint * Rename test file * Add block tests, add cheap block proposer check * Rename block testing file * Add observed_block_producers * Tidy * Switch around block signature verification * Finish block testing * Remove gossip from signature tests * First pass of self review * Fix deviation in spec * Update test spec tags * Start moving over to hashset * Finish moving observed attesters to hashmap * Move aggregation pool over to hashmap * Make fc attn borrow again * Fix rest_api compile error * Fix missing comments * Fix monster test * Uncomment increasing slots test * Address remaining comments * Remove unsafe, use cfg test * Remove cfg test flag * Fix dodgy comment * Ignore aggregates that are already known. * Unify aggregator modulo logic * Fix typo in logs * Refactor validator subscription logic * Avoid reproducing selection proof * Skip HTTP call if no subscriptions * Rename DutyAndState -> DutyAndProof * Tidy logs * Print root as dbg * Fix compile errors in tests * Fix compile error in test
2020-05-06 11:42:56 +00:00
let current_slot = harness.chain.slot().expect("should get slot");
let expected_attestation_slot = attestation.data.slot;
Add attestation gossip pre-verification (#983) * Add PH & MS slot clock changes * Account for genesis time * Add progress on duties refactor * Add simple is_aggregator bool to val subscription * Start work on attestation_verification.rs * Add progress on ObservedAttestations * Progress with ObservedAttestations * Fix tests * Add observed attestations to the beacon chain * Add attestation observation to processing code * Add progress on attestation verification * Add first draft of ObservedAttesters * Add more tests * Add observed attesters to beacon chain * Add observers to attestation processing * Add more attestation verification * Create ObservedAggregators map * Remove commented-out code * Add observed aggregators into chain * Add progress * Finish adding features to attestation verification * Ensure beacon chain compiles * Link attn verification into chain * Integrate new attn verification in chain * Remove old attestation processing code * Start trying to fix beacon_chain tests * Split adding into pools into two functions * Add aggregation to harness * Get test harness working again * Adjust the number of aggregators for test harness * Fix edge-case in harness * Integrate new attn processing in network * Fix compile bug in validator_client * Update validator API endpoints * Fix aggreagation in test harness * Fix enum thing * Fix attestation observation bug: * Patch failing API tests * Start adding comments to attestation verification * Remove unused attestation field * Unify "is block known" logic * Update comments * Supress fork choice errors for network processing * Add todos * Tidy * Add gossip attn tests * Disallow test harness to produce old attns * Comment out in-progress tests * Partially address pruning tests * Fix failing store test * Add aggregate tests * Add comments about which spec conditions we check * Dont re-aggregate * Split apart test harness attn production * Fix compile error in network * Make progress on commented-out test * Fix skipping attestation test * Add fork choice verification tests * Tidy attn tests, remove dead code * Remove some accidentally added code * Fix clippy lint * Rename test file * Add block tests, add cheap block proposer check * Rename block testing file * Add observed_block_producers * Tidy * Switch around block signature verification * Finish block testing * Remove gossip from signature tests * First pass of self review * Fix deviation in spec * Update test spec tags * Start moving over to hashset * Finish moving observed attesters to hashmap * Move aggregation pool over to hashmap * Make fc attn borrow again * Fix rest_api compile error * Fix missing comments * Fix monster test * Uncomment increasing slots test * Address remaining comments * Remove unsafe, use cfg test * Remove cfg test flag * Fix dodgy comment * Ignore aggregates that are already known. * Unify aggregator modulo logic * Fix typo in logs * Refactor validator subscription logic * Avoid reproducing selection proof * Skip HTTP call if no subscriptions * Rename DutyAndState -> DutyAndProof * Tidy logs * Print root as dbg * Fix compile errors in tests * Fix compile error in test
2020-05-06 11:42:56 +00:00
// Extra -1 to handle gossip clock disparity.
let expected_earliest_permissible_slot = current_slot - E::slots_per_epoch() - 1;
Optimize attestation processing (#841) * Start updating types * WIP * Signature hacking * Existing EF tests passing with fake_crypto * Updates * Delete outdated API spec * The refactor continues * It compiles * WIP test fixes * All release tests passing bar genesis state parsing * Update and test YamlConfig * Update to spec v0.10 compatible BLS * Updates to BLS EF tests * Add EF test for AggregateVerify And delete unused hash2curve tests for uncompressed points * Update EF tests to v0.10.1 * Use optional block root correctly in block proc * Use genesis fork in deposit domain. All tests pass * Cargo fmt * Fast aggregate verify test * Update REST API docs * Cargo fmt * Fix unused import * Bump spec tags to v0.10.1 * Add `seconds_per_eth1_block` to chainspec * Update to timestamp based eth1 voting scheme * Return None from `get_votes_to_consider` if block cache is empty * Handle overflows in `is_candidate_block` * Revert to failing tests * Fix eth1 data sets test * Choose default vote according to spec * Fix collect_valid_votes tests * Fix `get_votes_to_consider` to choose all eligible blocks * Uncomment winning_vote tests * Add comments; remove unused code * Reduce seconds_per_eth1_block for simulation * Addressed review comments * Add test for default vote case * Fix logs * Remove unused functions * Meter default eth1 votes * Fix comments * Address review comments; remove unused dependency * Add first attempt at attestation proc. re-write * Add version 2 of attestation processing * Minor fixes * Add validator pubkey cache * Make get_indexed_attestation take a committee * Link signature processing into new attn verification * First working version * Ensure pubkey cache is updated * Add more metrics, slight optimizations * Clone committee cache during attestation processing * Update shuffling cache during block processing * Remove old commented-out code * Fix shuffling cache insert bug * Used indexed attestation in fork choice * Restructure attn processing, add metrics * Add more detailed metrics * Tidy, fix failing tests * Fix failing tests, tidy * Disable/delete two outdated tests * Tidy * Add pubkey cache persistence file * Add more comments * Integrate persistence file into builder * Add pubkey cache tests * Add data_dir to beacon chain builder * Remove Option in pubkey cache persistence file * Ensure consistency between datadir/data_dir * Fix failing network test * Tidy * Fix todos * Add attestation processing tests * Add another test * Only run attestation tests in release * Make attestation tests MainnetEthSpec * Address Michael's comments * Remove redundant check * Fix warning * Fix failing test Co-authored-by: Michael Sproul <micsproul@gmail.com> Co-authored-by: Pawan Dhananjay <pawandhananjay@gmail.com>
2020-03-05 06:19:35 +00:00
if expected_attestation_slot <= finalized_epoch.start_slot(E::slots_per_epoch())
|| expected_attestation_slot < expected_earliest_permissible_slot
Optimize attestation processing (#841) * Start updating types * WIP * Signature hacking * Existing EF tests passing with fake_crypto * Updates * Delete outdated API spec * The refactor continues * It compiles * WIP test fixes * All release tests passing bar genesis state parsing * Update and test YamlConfig * Update to spec v0.10 compatible BLS * Updates to BLS EF tests * Add EF test for AggregateVerify And delete unused hash2curve tests for uncompressed points * Update EF tests to v0.10.1 * Use optional block root correctly in block proc * Use genesis fork in deposit domain. All tests pass * Cargo fmt * Fast aggregate verify test * Update REST API docs * Cargo fmt * Fix unused import * Bump spec tags to v0.10.1 * Add `seconds_per_eth1_block` to chainspec * Update to timestamp based eth1 voting scheme * Return None from `get_votes_to_consider` if block cache is empty * Handle overflows in `is_candidate_block` * Revert to failing tests * Fix eth1 data sets test * Choose default vote according to spec * Fix collect_valid_votes tests * Fix `get_votes_to_consider` to choose all eligible blocks * Uncomment winning_vote tests * Add comments; remove unused code * Reduce seconds_per_eth1_block for simulation * Addressed review comments * Add test for default vote case * Fix logs * Remove unused functions * Meter default eth1 votes * Fix comments * Address review comments; remove unused dependency * Add first attempt at attestation proc. re-write * Add version 2 of attestation processing * Minor fixes * Add validator pubkey cache * Make get_indexed_attestation take a committee * Link signature processing into new attn verification * First working version * Ensure pubkey cache is updated * Add more metrics, slight optimizations * Clone committee cache during attestation processing * Update shuffling cache during block processing * Remove old commented-out code * Fix shuffling cache insert bug * Used indexed attestation in fork choice * Restructure attn processing, add metrics * Add more detailed metrics * Tidy, fix failing tests * Fix failing tests, tidy * Disable/delete two outdated tests * Tidy * Add pubkey cache persistence file * Add more comments * Integrate persistence file into builder * Add pubkey cache tests * Add data_dir to beacon chain builder * Remove Option in pubkey cache persistence file * Ensure consistency between datadir/data_dir * Fix failing network test * Tidy * Fix todos * Add attestation processing tests * Add another test * Only run attestation tests in release * Make attestation tests MainnetEthSpec * Address Michael's comments * Remove redundant check * Fix warning * Fix failing test Co-authored-by: Michael Sproul <micsproul@gmail.com> Co-authored-by: Pawan Dhananjay <pawandhananjay@gmail.com>
2020-03-05 06:19:35 +00:00
{
checked_pre_fin = true;
assert!(matches!(
Batch BLS verification for attestations (#2399) ## Issue Addressed NA ## Proposed Changes Adds the ability to verify batches of aggregated/unaggregated attestations from the network. When the `BeaconProcessor` finds there are messages in the aggregated or unaggregated attestation queues, it will first check the length of the queue: - `== 1` verify the attestation individually. - `>= 2` take up to 64 of those attestations and verify them in a batch. Notably, we only perform batch verification if the queue has a backlog. We don't apply any artificial delays to attestations to try and force them into batches. ### Batching Details To assist with implementing batches we modify `beacon_chain::attestation_verification` to have two distinct categories for attestations: - *Indexed* attestations: those which have passed initial validation and were valid enough for us to derive an `IndexedAttestation`. - *Verified* attestations: those attestations which were indexed *and also* passed signature verification. These are well-formed, interesting messages which were signed by validators. The batching functions accept `n` attestations and then return `n` attestation verification `Result`s, where those `Result`s can be any combination of `Ok` or `Err`. In other words, we attempt to verify as many attestations as possible and return specific per-attestation results so peer scores can be updated, if required. When we batch verify attestations, we first try to map all those attestations to *indexed* attestations. If any of those attestations were able to be indexed, we then perform batch BLS verification on those indexed attestations. If the batch verification succeeds, we convert them into *verified* attestations, disabling individual signature checking. If the batch fails, we convert to verified attestations with individual signature checking enabled. Ultimately, we optimistically try to do a batch verification of attestation signatures and fall-back to individual verification if it fails. This opens an attach vector for "poisoning" the attestations and causing us to waste a batch verification. I argue that peer scoring should do a good-enough job of defending against this and the typical-case gains massively outweigh the worst-case losses. ## Additional Info Before this PR, attestation verification took the attestations by value (instead of by reference). It turns out that this was unnecessary and, in my opinion, resulted in some undesirable ergonomics (e.g., we had to pass the attestation back in the `Err` variant to avoid clones). In this PR I've modified attestation verification so that it now takes a reference. I refactored the `beacon_chain/tests/attestation_verification.rs` tests so they use a builder-esque "tester" struct instead of a weird macro. It made it easier for me to test individual/batch with the same set of tests and I think it was a nice tidy-up. Notably, I did this last to try and make sure my new refactors to *actual* production code would pass under the existing test suite.
2021-09-22 08:49:41 +00:00
res.err().unwrap(),
Add attestation gossip pre-verification (#983) * Add PH & MS slot clock changes * Account for genesis time * Add progress on duties refactor * Add simple is_aggregator bool to val subscription * Start work on attestation_verification.rs * Add progress on ObservedAttestations * Progress with ObservedAttestations * Fix tests * Add observed attestations to the beacon chain * Add attestation observation to processing code * Add progress on attestation verification * Add first draft of ObservedAttesters * Add more tests * Add observed attesters to beacon chain * Add observers to attestation processing * Add more attestation verification * Create ObservedAggregators map * Remove commented-out code * Add observed aggregators into chain * Add progress * Finish adding features to attestation verification * Ensure beacon chain compiles * Link attn verification into chain * Integrate new attn verification in chain * Remove old attestation processing code * Start trying to fix beacon_chain tests * Split adding into pools into two functions * Add aggregation to harness * Get test harness working again * Adjust the number of aggregators for test harness * Fix edge-case in harness * Integrate new attn processing in network * Fix compile bug in validator_client * Update validator API endpoints * Fix aggreagation in test harness * Fix enum thing * Fix attestation observation bug: * Patch failing API tests * Start adding comments to attestation verification * Remove unused attestation field * Unify "is block known" logic * Update comments * Supress fork choice errors for network processing * Add todos * Tidy * Add gossip attn tests * Disallow test harness to produce old attns * Comment out in-progress tests * Partially address pruning tests * Fix failing store test * Add aggregate tests * Add comments about which spec conditions we check * Dont re-aggregate * Split apart test harness attn production * Fix compile error in network * Make progress on commented-out test * Fix skipping attestation test * Add fork choice verification tests * Tidy attn tests, remove dead code * Remove some accidentally added code * Fix clippy lint * Rename test file * Add block tests, add cheap block proposer check * Rename block testing file * Add observed_block_producers * Tidy * Switch around block signature verification * Finish block testing * Remove gossip from signature tests * First pass of self review * Fix deviation in spec * Update test spec tags * Start moving over to hashset * Finish moving observed attesters to hashmap * Move aggregation pool over to hashmap * Make fc attn borrow again * Fix rest_api compile error * Fix missing comments * Fix monster test * Uncomment increasing slots test * Address remaining comments * Remove unsafe, use cfg test * Remove cfg test flag * Fix dodgy comment * Ignore aggregates that are already known. * Unify aggregator modulo logic * Fix typo in logs * Refactor validator subscription logic * Avoid reproducing selection proof * Skip HTTP call if no subscriptions * Rename DutyAndState -> DutyAndProof * Tidy logs * Print root as dbg * Fix compile errors in tests * Fix compile error in test
2020-05-06 11:42:56 +00:00
AttnError::PastSlot {
attestation_slot,
earliest_permissible_slot,
}
if attestation_slot == expected_attestation_slot && earliest_permissible_slot == expected_earliest_permissible_slot
));
} else {
Add attestation gossip pre-verification (#983) * Add PH & MS slot clock changes * Account for genesis time * Add progress on duties refactor * Add simple is_aggregator bool to val subscription * Start work on attestation_verification.rs * Add progress on ObservedAttestations * Progress with ObservedAttestations * Fix tests * Add observed attestations to the beacon chain * Add attestation observation to processing code * Add progress on attestation verification * Add first draft of ObservedAttesters * Add more tests * Add observed attesters to beacon chain * Add observers to attestation processing * Add more attestation verification * Create ObservedAggregators map * Remove commented-out code * Add observed aggregators into chain * Add progress * Finish adding features to attestation verification * Ensure beacon chain compiles * Link attn verification into chain * Integrate new attn verification in chain * Remove old attestation processing code * Start trying to fix beacon_chain tests * Split adding into pools into two functions * Add aggregation to harness * Get test harness working again * Adjust the number of aggregators for test harness * Fix edge-case in harness * Integrate new attn processing in network * Fix compile bug in validator_client * Update validator API endpoints * Fix aggreagation in test harness * Fix enum thing * Fix attestation observation bug: * Patch failing API tests * Start adding comments to attestation verification * Remove unused attestation field * Unify "is block known" logic * Update comments * Supress fork choice errors for network processing * Add todos * Tidy * Add gossip attn tests * Disallow test harness to produce old attns * Comment out in-progress tests * Partially address pruning tests * Fix failing store test * Add aggregate tests * Add comments about which spec conditions we check * Dont re-aggregate * Split apart test harness attn production * Fix compile error in network * Make progress on commented-out test * Fix skipping attestation test * Add fork choice verification tests * Tidy attn tests, remove dead code * Remove some accidentally added code * Fix clippy lint * Rename test file * Add block tests, add cheap block proposer check * Rename block testing file * Add observed_block_producers * Tidy * Switch around block signature verification * Finish block testing * Remove gossip from signature tests * First pass of self review * Fix deviation in spec * Update test spec tags * Start moving over to hashset * Finish moving observed attesters to hashmap * Move aggregation pool over to hashmap * Make fc attn borrow again * Fix rest_api compile error * Fix missing comments * Fix monster test * Uncomment increasing slots test * Address remaining comments * Remove unsafe, use cfg test * Remove cfg test flag * Fix dodgy comment * Ignore aggregates that are already known. * Unify aggregator modulo logic * Fix typo in logs * Refactor validator subscription logic * Avoid reproducing selection proof * Skip HTTP call if no subscriptions * Rename DutyAndState -> DutyAndProof * Tidy logs * Print root as dbg * Fix compile errors in tests * Fix compile error in test
2020-05-06 11:42:56 +00:00
res.expect("should have verified attetation");
}
}
assert!(checked_pre_fin);
}
// Test that the `end_slot` for forwards block and state root iterators works correctly.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn forwards_iter_block_and_state_roots_until() {
let num_blocks_produced = E::slots_per_epoch() * 17;
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
let harness = get_harness(store.clone(), LOW_VALIDATOR_COUNT);
let all_validators = &harness.get_all_validators();
let (mut head_state, mut head_state_root) = harness.get_current_state_and_root();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let head_block_root = harness.head_block_root();
let mut block_roots = vec![head_block_root];
let mut state_roots = vec![head_state_root];
for slot in (1..=num_blocks_produced).map(Slot::from) {
let (block_root, mut state) = harness
.add_attested_block_at_slot(slot, head_state, head_state_root, all_validators)
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
.await
.unwrap();
head_state_root = state.update_tree_hash_cache().unwrap();
head_state = state;
block_roots.push(block_root.into());
state_roots.push(head_state_root);
}
check_finalization(&harness, num_blocks_produced);
check_split_slot(&harness, store.clone());
// The last restore point slot is the point at which the hybrid forwards iterator behaviour
// changes.
let last_restore_point_slot = store.get_latest_restore_point_slot();
assert!(last_restore_point_slot > 0);
let chain = &harness.chain;
let head_state = harness.get_current_state();
let head_slot = head_state.slot();
assert_eq!(head_slot, num_blocks_produced);
let test_range = |start_slot: Slot, end_slot: Slot| {
let mut block_root_iter = chain
.forwards_iter_block_roots_until(start_slot, end_slot)
.unwrap();
let mut state_root_iter = chain
.forwards_iter_state_roots_until(start_slot, end_slot)
.unwrap();
for slot in (start_slot.as_u64()..=end_slot.as_u64()).map(Slot::new) {
let block_root = block_roots[slot.as_usize()];
assert_eq!(block_root_iter.next().unwrap().unwrap(), (block_root, slot));
let state_root = state_roots[slot.as_usize()];
assert_eq!(state_root_iter.next().unwrap().unwrap(), (state_root, slot));
}
};
let split_slot = store.get_split_slot();
assert!(split_slot > last_restore_point_slot);
test_range(Slot::new(0), last_restore_point_slot);
test_range(last_restore_point_slot, last_restore_point_slot);
test_range(last_restore_point_slot - 1, last_restore_point_slot);
test_range(Slot::new(0), last_restore_point_slot - 1);
test_range(Slot::new(0), split_slot);
test_range(last_restore_point_slot - 1, split_slot);
test_range(Slot::new(0), head_state.slot());
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn block_replay_with_inaccurate_state_roots() {
let num_blocks_produced = E::slots_per_epoch() * 3 + 31;
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
let harness = get_harness(store.clone(), LOW_VALIDATOR_COUNT);
let chain = &harness.chain;
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
num_blocks_produced as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
// Slot must not be 0 mod 32 or else no blocks will be replayed.
let (mut head_state, head_root) = harness.get_current_state_and_root();
assert_ne!(head_state.slot() % 32, 0);
let mut fast_head_state = store
.get_inconsistent_state_for_attestation_verification_only(
&head_root,
Some(head_state.slot()),
)
.unwrap()
.unwrap();
assert_eq!(head_state.validators(), fast_head_state.validators());
head_state.build_all_committee_caches(&chain.spec).unwrap();
fast_head_state
.build_all_committee_caches(&chain.spec)
.unwrap();
assert_eq!(
head_state
.get_cached_active_validator_indices(RelativeEpoch::Current)
.unwrap(),
fast_head_state
.get_cached_active_validator_indices(RelativeEpoch::Current)
.unwrap()
);
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn block_replayer_hooks() {
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
let harness = get_harness(store.clone(), LOW_VALIDATOR_COUNT);
let chain = &harness.chain;
let block_slots = vec![1, 3, 5, 10, 11, 12, 13, 14, 31, 32, 33]
.into_iter()
.map(Slot::new)
.collect::<Vec<_>>();
let max_slot = *block_slots.last().unwrap();
let all_slots = (0..=max_slot.as_u64()).map(Slot::new).collect::<Vec<_>>();
let (state, state_root) = harness.get_current_state_and_root();
let all_validators = harness.get_all_validators();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (_, _, end_block_root, mut end_state) = harness
.add_attested_blocks_at_slots(state.clone(), state_root, &block_slots, &all_validators)
.await;
let blocks = store
.load_blocks_to_replay(Slot::new(0), max_slot, end_block_root.into())
.unwrap();
let mut pre_slots = vec![];
let mut post_slots = vec![];
let mut pre_block_slots = vec![];
let mut post_block_slots = vec![];
let mut replay_state = BlockReplayer::<MinimalEthSpec>::new(state, &chain.spec)
.pre_slot_hook(Box::new(|state| {
pre_slots.push(state.slot());
Ok(())
}))
.post_slot_hook(Box::new(|state, epoch_summary, is_skip_slot| {
if is_skip_slot {
assert!(!block_slots.contains(&state.slot()));
} else {
assert!(block_slots.contains(&state.slot()));
}
if state.slot() % E::slots_per_epoch() == 0 {
assert!(epoch_summary.is_some());
}
post_slots.push(state.slot());
Ok(())
}))
.pre_block_hook(Box::new(|state, block| {
assert_eq!(state.slot(), block.slot());
pre_block_slots.push(block.slot());
Ok(())
}))
.post_block_hook(Box::new(|state, block| {
assert_eq!(state.slot(), block.slot());
post_block_slots.push(block.slot());
Ok(())
}))
.apply_blocks(blocks, None)
.unwrap()
.into_state();
// All but last slot seen by pre-slot hook
assert_eq!(&pre_slots, all_slots.split_last().unwrap().1);
// All but 0th slot seen by post-slot hook
assert_eq!(&post_slots, all_slots.split_first().unwrap().1);
// All blocks seen by both hooks
assert_eq!(pre_block_slots, block_slots);
assert_eq!(post_block_slots, block_slots);
// States match.
end_state.drop_all_caches().unwrap();
replay_state.drop_all_caches().unwrap();
assert_eq!(end_state, replay_state);
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn delete_blocks_and_states() {
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
let validators_keypairs =
types::test_utils::generate_deterministic_keypairs(LOW_VALIDATOR_COUNT);
let harness = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.keypairs(validators_keypairs)
.fresh_disk_store(store.clone())
.mock_execution_layer()
.build();
let unforked_blocks: u64 = 4 * E::slots_per_epoch();
// Finalize an initial portion of the chain.
let initial_slots: Vec<Slot> = (1..=unforked_blocks).map(Into::into).collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (state, state_root) = harness.get_current_state_and_root();
let all_validators = harness.get_all_validators();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.add_attested_blocks_at_slots(state, state_root, &initial_slots, &all_validators)
.await;
// Create a fork post-finalization.
let two_thirds = (LOW_VALIDATOR_COUNT / 3) * 2;
let honest_validators: Vec<usize> = (0..two_thirds).collect();
let faulty_validators: Vec<usize> = (two_thirds..LOW_VALIDATOR_COUNT).collect();
let fork_blocks = 2 * E::slots_per_epoch();
let slot_u64: u64 = harness.get_current_slot().as_u64() + 1;
let fork1_slots: Vec<Slot> = (slot_u64..(slot_u64 + fork_blocks))
.map(Into::into)
.collect();
let fork2_slots: Vec<Slot> = (slot_u64 + 1..(slot_u64 + 1 + fork_blocks))
.map(Into::into)
.collect();
let fork1_state = harness.get_current_state();
let fork2_state = fork1_state.clone();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let results = harness
.add_blocks_on_multiple_chains(vec![
(fork1_state, fork1_slots, honest_validators),
(fork2_state, fork2_slots, faulty_validators),
])
.await;
let honest_head = results[0].2;
let faulty_head = results[1].2;
assert_ne!(honest_head, faulty_head, "forks should be distinct");
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
assert_eq!(harness.head_slot(), unforked_blocks + fork_blocks);
assert_eq!(
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness.head_block_root(),
honest_head.into(),
"the honest chain should be the canonical chain",
);
let faulty_head_block = store
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
.get_blinded_block(&faulty_head.into())
.expect("no errors")
.expect("faulty head block exists");
let faulty_head_state = store
.get_state(
&faulty_head_block.state_root(),
Some(faulty_head_block.slot()),
)
.expect("no db error")
.expect("faulty head state exists");
// Delete faulty fork
// Attempting to load those states should find them unavailable
for (state_root, slot) in
StateRootsIterator::new(&store, &faulty_head_state).map(Result::unwrap)
{
if slot <= unforked_blocks {
break;
}
store.delete_state(&state_root, slot).unwrap();
assert_eq!(store.get_state(&state_root, Some(slot)).unwrap(), None);
}
// Double-deleting should also be OK (deleting non-existent things is fine)
for (state_root, slot) in
StateRootsIterator::new(&store, &faulty_head_state).map(Result::unwrap)
{
if slot <= unforked_blocks {
break;
}
store.delete_state(&state_root, slot).unwrap();
}
// Deleting the blocks from the fork should remove them completely
for (block_root, slot) in
BlockRootsIterator::new(&store, &faulty_head_state).map(Result::unwrap)
{
if slot <= unforked_blocks + 1 {
break;
}
store.delete_block(&block_root).unwrap();
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
assert_eq!(store.get_blinded_block(&block_root).unwrap(), None);
}
// Deleting frozen states should do nothing
let split_slot = store.get_split_slot();
let finalized_states = harness
.chain
.forwards_iter_state_roots(Slot::new(0))
.expect("should get iter")
.map(Result::unwrap);
for (state_root, slot) in finalized_states {
if slot < split_slot {
store.delete_state(&state_root, slot).unwrap();
}
}
// After all that, the chain dump should still be OK
check_chain_dump(&harness, unforked_blocks + fork_blocks + 1);
}
// Check that we never produce invalid blocks when there is deep forking that changes the shuffling.
// See https://github.com/sigp/lighthouse/issues/845
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
async fn multi_epoch_fork_valid_blocks_test(
initial_blocks: usize,
num_fork1_blocks_: usize,
num_fork2_blocks_: usize,
num_fork1_validators: usize,
) -> (TempDir, TestHarness, Hash256, Hash256) {
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
let validators_keypairs =
types::test_utils::generate_deterministic_keypairs(LOW_VALIDATOR_COUNT);
let harness = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.keypairs(validators_keypairs)
.fresh_disk_store(store)
.mock_execution_layer()
.build();
let num_fork1_blocks: u64 = num_fork1_blocks_.try_into().unwrap();
let num_fork2_blocks: u64 = num_fork2_blocks_.try_into().unwrap();
// Create the initial portion of the chain
if initial_blocks > 0 {
let initial_slots: Vec<Slot> = (1..=initial_blocks).map(Into::into).collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (state, state_root) = harness.get_current_state_and_root();
let all_validators = harness.get_all_validators();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.add_attested_blocks_at_slots(state, state_root, &initial_slots, &all_validators)
.await;
}
assert!(num_fork1_validators <= LOW_VALIDATOR_COUNT);
let fork1_validators: Vec<usize> = (0..num_fork1_validators).collect();
let fork2_validators: Vec<usize> = (num_fork1_validators..LOW_VALIDATOR_COUNT).collect();
let fork1_state = harness.get_current_state();
let fork2_state = fork1_state.clone();
let slot_u64: u64 = harness.get_current_slot().as_u64() + 1;
let fork1_slots: Vec<Slot> = (slot_u64..(slot_u64 + num_fork1_blocks))
.map(Into::into)
.collect();
let fork2_slots: Vec<Slot> = (slot_u64 + 1..(slot_u64 + 1 + num_fork2_blocks))
.map(Into::into)
.collect();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let results = harness
.add_blocks_on_multiple_chains(vec![
(fork1_state, fork1_slots, fork1_validators),
(fork2_state, fork2_slots, fork2_validators),
])
.await;
let head1 = results[0].2;
let head2 = results[1].2;
(db_path, harness, head1.into(), head2.into())
}
// This is the minimal test of block production with different shufflings.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn block_production_different_shuffling_early() {
let slots_per_epoch = E::slots_per_epoch() as usize;
multi_epoch_fork_valid_blocks_test(
slots_per_epoch - 2,
slots_per_epoch + 3,
slots_per_epoch + 3,
LOW_VALIDATOR_COUNT / 2,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn block_production_different_shuffling_long() {
let slots_per_epoch = E::slots_per_epoch() as usize;
multi_epoch_fork_valid_blocks_test(
2 * slots_per_epoch - 2,
3 * slots_per_epoch,
3 * slots_per_epoch,
LOW_VALIDATOR_COUNT / 2,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
}
// Check that the op pool safely includes multiple attestations per block when necessary.
// This checks the correctness of the shuffling compatibility memoization.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn multiple_attestations_per_block() {
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
let harness = get_harness(store, HIGH_VALIDATOR_COUNT);
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
E::slots_per_epoch() as usize * 3,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let head = harness.chain.head_snapshot();
let committees_per_slot = head
.beacon_state
.get_committee_count_at_slot(head.beacon_state.slot())
.unwrap();
assert!(committees_per_slot > 1);
for snapshot in harness.chain.chain_dump().unwrap() {
let slot = snapshot.beacon_block.slot();
assert_eq!(
snapshot
.beacon_block
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
.as_ref()
Refactor op pool for speed and correctness (#3312) ## Proposed Changes This PR has two aims: to speed up attestation packing in the op pool, and to fix bugs in the verification of attester slashings, proposer slashings and voluntary exits. The changes are bundled into a single database schema upgrade (v12). Attestation packing is sped up by removing several inefficiencies: - No more recalculation of `attesting_indices` during packing. - No (unnecessary) examination of the `ParticipationFlags`: a bitfield suffices. See `RewardCache`. - No re-checking of attestation validity during packing: the `AttestationMap` provides attestations which are "correct by construction" (I have checked this using Hydra). - No SSZ re-serialization for the clunky `AttestationId` type (it can be removed in a future release). So far the speed-up seems to be roughly 2-10x, from 500ms down to 50-100ms. Verification of attester slashings, proposer slashings and voluntary exits is fixed by: - Tracking the `ForkVersion`s that were used to verify each message inside the `SigVerifiedOp`. This allows us to quickly re-verify that they match the head state's opinion of what the `ForkVersion` should be at the epoch(s) relevant to the message. - Storing the `SigVerifiedOp` on disk rather than the raw operation. This allows us to continue track the fork versions after a reboot. This is mostly contained in this commit 52bb1840ae5c4356a8fc3a51e5df23ed65ed2c7f. ## Additional Info The schema upgrade uses the justified state to re-verify attestations and compute `attesting_indices` for them. It will drop any attestations that fail to verify, by the logic that attestations are most valuable in the few slots after they're observed, and are probably stale and useless by the time a node restarts. Exits and proposer slashings and similarly re-verified to obtain `SigVerifiedOp`s. This PR contains a runtime killswitch `--paranoid-block-proposal` which opts out of all the optimisations in favour of closely verifying every included message. Although I'm quite sure that the optimisations are correct this flag could be useful in the event of an unforeseen emergency. Finally, you might notice that the `RewardCache` appears quite useless in its current form because it is only updated on the hot-path immediately before proposal. My hope is that in future we can shift calls to `RewardCache::update` into the background, e.g. while performing the state advance. It is also forward-looking to `tree-states` compatibility, where iterating and indexing `state.{previous,current}_epoch_participation` is expensive and needs to be minimised.
2022-08-29 09:10:26 +00:00
.message()
.body()
.attestations()
.len() as u64,
if slot <= 1 { 0 } else { committees_per_slot }
);
}
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn shuffling_compatible_linear_chain() {
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
let harness = get_harness(store.clone(), LOW_VALIDATOR_COUNT);
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let head_block_root = harness
.extend_chain(
4 * E::slots_per_epoch() as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
check_shuffling_compatible(
&harness,
&get_state_for_block(&harness, head_block_root),
head_block_root,
);
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn shuffling_compatible_missing_pivot_block() {
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
let harness = get_harness(store.clone(), LOW_VALIDATOR_COUNT);
// Skip the block at the end of the first epoch.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
E::slots_per_epoch() as usize - 2,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
harness.advance_slot();
harness.advance_slot();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let head_block_root = harness
.extend_chain(
2 * E::slots_per_epoch() as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
check_shuffling_compatible(
&harness,
&get_state_for_block(&harness, head_block_root),
head_block_root,
);
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn shuffling_compatible_simple_fork() {
let slots_per_epoch = E::slots_per_epoch() as usize;
let (db_path, harness, head1, head2) = multi_epoch_fork_valid_blocks_test(
2 * slots_per_epoch,
3 * slots_per_epoch,
3 * slots_per_epoch,
LOW_VALIDATOR_COUNT / 2,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
let head1_state = get_state_for_block(&harness, head1);
let head2_state = get_state_for_block(&harness, head2);
Fix attestation shuffling filter (#3629) ## Issue Addressed Fix a bug in block production that results in blocks with 0 attestations during the first slot of an epoch. The bug is marked by debug logs of the form: > DEBG Discarding attestation because of missing ancestor, block_root: 0x3cc00d9c9e0883b2d0db8606278f2b8423d4902f9a1ee619258b5b60590e64f8, pivot_slot: 4042591 It occurs when trying to look up the shuffling decision root for an attestation from a slot which is prior to fork choice's finalized block. This happens frequently when proposing in the first slot of the epoch where we have: - `current_epoch == n` - `attestation.data.target.epoch == n - 1` - attestation shuffling epoch `== n - 3` (decision block being the last block of `n - 3`) - `state.finalized_checkpoint.epoch == n - 2` (first block of `n - 2` is finalized) Hence the shuffling decision slot is out of range of the fork choice backwards iterator _by a single slot_. Unfortunately this bug was hidden when we weren't pruning fork choice, and then reintroduced in v2.5.1 when we fixed the pruning (https://github.com/sigp/lighthouse/releases/tag/v2.5.1). There's no way to turn that off or disable the filtering in our current release, so we need a new release to fix this issue. Fortunately, it also does not occur on every epoch boundary because of the gradual pruning of fork choice every 256 blocks (~8 epochs): https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array_fork_choice.rs#L16 https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array.rs#L713-L716 So the probability of proposing a 0-attestation block given a proposal assignment is approximately `1/32 * 1/8 = 0.39%`. ## Proposed Changes - Load the block's shuffling ID from fork choice and verify it against the expected shuffling ID of the head state. This code was initially written before we had settled on a representation of shuffling IDs, so I think it's a nice simplification to make use of them here rather than more ad-hoc logic that fundamentally does the same thing. ## Additional Info Thanks to @moshe-blox for noticing this issue and bringing it to our attention.
2022-10-18 04:02:06 +00:00
check_shuffling_compatible(&harness, &head1_state, head1);
check_shuffling_compatible(&harness, &head1_state, head2);
check_shuffling_compatible(&harness, &head2_state, head1);
check_shuffling_compatible(&harness, &head2_state, head2);
drop(db_path);
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn shuffling_compatible_short_fork() {
let slots_per_epoch = E::slots_per_epoch() as usize;
let (db_path, harness, head1, head2) = multi_epoch_fork_valid_blocks_test(
2 * slots_per_epoch - 2,
slots_per_epoch + 2,
slots_per_epoch + 2,
LOW_VALIDATOR_COUNT / 2,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
let head1_state = get_state_for_block(&harness, head1);
let head2_state = get_state_for_block(&harness, head2);
Fix attestation shuffling filter (#3629) ## Issue Addressed Fix a bug in block production that results in blocks with 0 attestations during the first slot of an epoch. The bug is marked by debug logs of the form: > DEBG Discarding attestation because of missing ancestor, block_root: 0x3cc00d9c9e0883b2d0db8606278f2b8423d4902f9a1ee619258b5b60590e64f8, pivot_slot: 4042591 It occurs when trying to look up the shuffling decision root for an attestation from a slot which is prior to fork choice's finalized block. This happens frequently when proposing in the first slot of the epoch where we have: - `current_epoch == n` - `attestation.data.target.epoch == n - 1` - attestation shuffling epoch `== n - 3` (decision block being the last block of `n - 3`) - `state.finalized_checkpoint.epoch == n - 2` (first block of `n - 2` is finalized) Hence the shuffling decision slot is out of range of the fork choice backwards iterator _by a single slot_. Unfortunately this bug was hidden when we weren't pruning fork choice, and then reintroduced in v2.5.1 when we fixed the pruning (https://github.com/sigp/lighthouse/releases/tag/v2.5.1). There's no way to turn that off or disable the filtering in our current release, so we need a new release to fix this issue. Fortunately, it also does not occur on every epoch boundary because of the gradual pruning of fork choice every 256 blocks (~8 epochs): https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array_fork_choice.rs#L16 https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array.rs#L713-L716 So the probability of proposing a 0-attestation block given a proposal assignment is approximately `1/32 * 1/8 = 0.39%`. ## Proposed Changes - Load the block's shuffling ID from fork choice and verify it against the expected shuffling ID of the head state. This code was initially written before we had settled on a representation of shuffling IDs, so I think it's a nice simplification to make use of them here rather than more ad-hoc logic that fundamentally does the same thing. ## Additional Info Thanks to @moshe-blox for noticing this issue and bringing it to our attention.
2022-10-18 04:02:06 +00:00
check_shuffling_compatible(&harness, &head1_state, head1);
check_shuffling_compatible(&harness, &head1_state, head2);
check_shuffling_compatible(&harness, &head2_state, head1);
check_shuffling_compatible(&harness, &head2_state, head2);
drop(db_path);
}
fn get_state_for_block(harness: &TestHarness, block_root: Hash256) -> BeaconState<E> {
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
let head_block = harness
.chain
.store
.get_blinded_block(&block_root)
.unwrap()
.unwrap();
harness
.chain
.get_state(&head_block.state_root(), Some(head_block.slot()))
.unwrap()
.unwrap()
}
/// Check the invariants that apply to `shuffling_is_compatible`.
fn check_shuffling_compatible(
harness: &TestHarness,
head_state: &BeaconState<E>,
head_block_root: Hash256,
) {
for maybe_tuple in harness
.chain
.rev_iter_block_roots_from(head_block_root)
.unwrap()
{
let (block_root, slot) = maybe_tuple.unwrap();
Fix attestation shuffling filter (#3629) ## Issue Addressed Fix a bug in block production that results in blocks with 0 attestations during the first slot of an epoch. The bug is marked by debug logs of the form: > DEBG Discarding attestation because of missing ancestor, block_root: 0x3cc00d9c9e0883b2d0db8606278f2b8423d4902f9a1ee619258b5b60590e64f8, pivot_slot: 4042591 It occurs when trying to look up the shuffling decision root for an attestation from a slot which is prior to fork choice's finalized block. This happens frequently when proposing in the first slot of the epoch where we have: - `current_epoch == n` - `attestation.data.target.epoch == n - 1` - attestation shuffling epoch `== n - 3` (decision block being the last block of `n - 3`) - `state.finalized_checkpoint.epoch == n - 2` (first block of `n - 2` is finalized) Hence the shuffling decision slot is out of range of the fork choice backwards iterator _by a single slot_. Unfortunately this bug was hidden when we weren't pruning fork choice, and then reintroduced in v2.5.1 when we fixed the pruning (https://github.com/sigp/lighthouse/releases/tag/v2.5.1). There's no way to turn that off or disable the filtering in our current release, so we need a new release to fix this issue. Fortunately, it also does not occur on every epoch boundary because of the gradual pruning of fork choice every 256 blocks (~8 epochs): https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array_fork_choice.rs#L16 https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array.rs#L713-L716 So the probability of proposing a 0-attestation block given a proposal assignment is approximately `1/32 * 1/8 = 0.39%`. ## Proposed Changes - Load the block's shuffling ID from fork choice and verify it against the expected shuffling ID of the head state. This code was initially written before we had settled on a representation of shuffling IDs, so I think it's a nice simplification to make use of them here rather than more ad-hoc logic that fundamentally does the same thing. ## Additional Info Thanks to @moshe-blox for noticing this issue and bringing it to our attention.
2022-10-18 04:02:06 +00:00
// Would an attestation to `block_root` at the current epoch be compatible with the head
// state's shuffling?
let current_epoch_shuffling_is_compatible = harness.chain.shuffling_is_compatible(
&block_root,
head_state.current_epoch(),
&head_state,
);
Fix attestation shuffling filter (#3629) ## Issue Addressed Fix a bug in block production that results in blocks with 0 attestations during the first slot of an epoch. The bug is marked by debug logs of the form: > DEBG Discarding attestation because of missing ancestor, block_root: 0x3cc00d9c9e0883b2d0db8606278f2b8423d4902f9a1ee619258b5b60590e64f8, pivot_slot: 4042591 It occurs when trying to look up the shuffling decision root for an attestation from a slot which is prior to fork choice's finalized block. This happens frequently when proposing in the first slot of the epoch where we have: - `current_epoch == n` - `attestation.data.target.epoch == n - 1` - attestation shuffling epoch `== n - 3` (decision block being the last block of `n - 3`) - `state.finalized_checkpoint.epoch == n - 2` (first block of `n - 2` is finalized) Hence the shuffling decision slot is out of range of the fork choice backwards iterator _by a single slot_. Unfortunately this bug was hidden when we weren't pruning fork choice, and then reintroduced in v2.5.1 when we fixed the pruning (https://github.com/sigp/lighthouse/releases/tag/v2.5.1). There's no way to turn that off or disable the filtering in our current release, so we need a new release to fix this issue. Fortunately, it also does not occur on every epoch boundary because of the gradual pruning of fork choice every 256 blocks (~8 epochs): https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array_fork_choice.rs#L16 https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array.rs#L713-L716 So the probability of proposing a 0-attestation block given a proposal assignment is approximately `1/32 * 1/8 = 0.39%`. ## Proposed Changes - Load the block's shuffling ID from fork choice and verify it against the expected shuffling ID of the head state. This code was initially written before we had settled on a representation of shuffling IDs, so I think it's a nice simplification to make use of them here rather than more ad-hoc logic that fundamentally does the same thing. ## Additional Info Thanks to @moshe-blox for noticing this issue and bringing it to our attention.
2022-10-18 04:02:06 +00:00
// Check for consistency with the more expensive shuffling lookup.
harness
.chain
.with_committee_cache(
block_root,
head_state.current_epoch(),
|committee_cache, _| {
let state_cache = head_state.committee_cache(RelativeEpoch::Current).unwrap();
if current_epoch_shuffling_is_compatible {
assert_eq!(committee_cache, state_cache, "block at slot {slot}");
} else {
assert_ne!(committee_cache, state_cache, "block at slot {slot}");
}
Ok(())
},
)
.unwrap_or_else(|e| {
// If the lookup fails then the shuffling must be invalid in some way, e.g. the
// block with `block_root` is from a later epoch than `previous_epoch`.
assert!(
!current_epoch_shuffling_is_compatible,
"block at slot {slot} has compatible shuffling at epoch {} \
but should be incompatible due to error: {e:?}",
head_state.current_epoch()
);
});
// Similarly for the previous epoch
Fix attestation shuffling filter (#3629) ## Issue Addressed Fix a bug in block production that results in blocks with 0 attestations during the first slot of an epoch. The bug is marked by debug logs of the form: > DEBG Discarding attestation because of missing ancestor, block_root: 0x3cc00d9c9e0883b2d0db8606278f2b8423d4902f9a1ee619258b5b60590e64f8, pivot_slot: 4042591 It occurs when trying to look up the shuffling decision root for an attestation from a slot which is prior to fork choice's finalized block. This happens frequently when proposing in the first slot of the epoch where we have: - `current_epoch == n` - `attestation.data.target.epoch == n - 1` - attestation shuffling epoch `== n - 3` (decision block being the last block of `n - 3`) - `state.finalized_checkpoint.epoch == n - 2` (first block of `n - 2` is finalized) Hence the shuffling decision slot is out of range of the fork choice backwards iterator _by a single slot_. Unfortunately this bug was hidden when we weren't pruning fork choice, and then reintroduced in v2.5.1 when we fixed the pruning (https://github.com/sigp/lighthouse/releases/tag/v2.5.1). There's no way to turn that off or disable the filtering in our current release, so we need a new release to fix this issue. Fortunately, it also does not occur on every epoch boundary because of the gradual pruning of fork choice every 256 blocks (~8 epochs): https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array_fork_choice.rs#L16 https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array.rs#L713-L716 So the probability of proposing a 0-attestation block given a proposal assignment is approximately `1/32 * 1/8 = 0.39%`. ## Proposed Changes - Load the block's shuffling ID from fork choice and verify it against the expected shuffling ID of the head state. This code was initially written before we had settled on a representation of shuffling IDs, so I think it's a nice simplification to make use of them here rather than more ad-hoc logic that fundamentally does the same thing. ## Additional Info Thanks to @moshe-blox for noticing this issue and bringing it to our attention.
2022-10-18 04:02:06 +00:00
let previous_epoch_shuffling_is_compatible = harness.chain.shuffling_is_compatible(
&block_root,
head_state.previous_epoch(),
&head_state,
);
Fix attestation shuffling filter (#3629) ## Issue Addressed Fix a bug in block production that results in blocks with 0 attestations during the first slot of an epoch. The bug is marked by debug logs of the form: > DEBG Discarding attestation because of missing ancestor, block_root: 0x3cc00d9c9e0883b2d0db8606278f2b8423d4902f9a1ee619258b5b60590e64f8, pivot_slot: 4042591 It occurs when trying to look up the shuffling decision root for an attestation from a slot which is prior to fork choice's finalized block. This happens frequently when proposing in the first slot of the epoch where we have: - `current_epoch == n` - `attestation.data.target.epoch == n - 1` - attestation shuffling epoch `== n - 3` (decision block being the last block of `n - 3`) - `state.finalized_checkpoint.epoch == n - 2` (first block of `n - 2` is finalized) Hence the shuffling decision slot is out of range of the fork choice backwards iterator _by a single slot_. Unfortunately this bug was hidden when we weren't pruning fork choice, and then reintroduced in v2.5.1 when we fixed the pruning (https://github.com/sigp/lighthouse/releases/tag/v2.5.1). There's no way to turn that off or disable the filtering in our current release, so we need a new release to fix this issue. Fortunately, it also does not occur on every epoch boundary because of the gradual pruning of fork choice every 256 blocks (~8 epochs): https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array_fork_choice.rs#L16 https://github.com/sigp/lighthouse/blob/01e84b71f524968f5b940fbd2fa31d29408b6581/consensus/proto_array/src/proto_array.rs#L713-L716 So the probability of proposing a 0-attestation block given a proposal assignment is approximately `1/32 * 1/8 = 0.39%`. ## Proposed Changes - Load the block's shuffling ID from fork choice and verify it against the expected shuffling ID of the head state. This code was initially written before we had settled on a representation of shuffling IDs, so I think it's a nice simplification to make use of them here rather than more ad-hoc logic that fundamentally does the same thing. ## Additional Info Thanks to @moshe-blox for noticing this issue and bringing it to our attention.
2022-10-18 04:02:06 +00:00
harness
.chain
.with_committee_cache(
block_root,
head_state.previous_epoch(),
|committee_cache, _| {
let state_cache = head_state.committee_cache(RelativeEpoch::Previous).unwrap();
if previous_epoch_shuffling_is_compatible {
assert_eq!(committee_cache, state_cache);
} else {
assert_ne!(committee_cache, state_cache);
}
Ok(())
},
)
.unwrap_or_else(|e| {
// If the lookup fails then the shuffling must be invalid in some way, e.g. the
// block with `block_root` is from a later epoch than `previous_epoch`.
assert!(
!previous_epoch_shuffling_is_compatible,
"block at slot {slot} has compatible shuffling at epoch {} \
but should be incompatible due to error: {e:?}",
head_state.previous_epoch()
);
});
// Targeting two epochs before the current epoch should always return false
if head_state.current_epoch() >= 2 {
assert_eq!(
harness.chain.shuffling_is_compatible(
&block_root,
head_state.current_epoch() - 2,
&head_state
),
false
);
}
}
}
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Ensure blocks from abandoned forks are pruned from the Hot DB
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn prunes_abandoned_fork_between_two_finalized_checkpoints() {
const HONEST_VALIDATOR_COUNT: usize = 16 + 0;
const ADVERSARIAL_VALIDATOR_COUNT: usize = 8 - 0;
const VALIDATOR_COUNT: usize = HONEST_VALIDATOR_COUNT + ADVERSARIAL_VALIDATOR_COUNT;
let validators_keypairs = types::test_utils::generate_deterministic_keypairs(VALIDATOR_COUNT);
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
let honest_validators: Vec<usize> = (0..HONEST_VALIDATOR_COUNT).collect();
let adversarial_validators: Vec<usize> = (HONEST_VALIDATOR_COUNT..VALIDATOR_COUNT).collect();
let rig = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.keypairs(validators_keypairs)
.fresh_ephemeral_store()
.mock_execution_layer()
.build();
let slots_per_epoch = rig.slots_per_epoch();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (mut state, state_root) = rig.get_current_state_and_root();
let canonical_chain_slots: Vec<Slot> = (1..=rig.epoch_start_slot(1)).map(Slot::new).collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (canonical_chain_blocks_pre_finalization, _, _, new_state) = rig
.add_attested_blocks_at_slots(
state,
state_root,
&canonical_chain_slots,
&honest_validators,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
state = new_state;
let canonical_chain_slot: u64 = rig.get_current_slot().into();
let stray_slots: Vec<Slot> = (canonical_chain_slot + 1..rig.epoch_start_slot(2))
.map(Slot::new)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (current_state, current_state_root) = rig.get_current_state_and_root();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (stray_blocks, stray_states, stray_head, _) = rig
.add_attested_blocks_at_slots(
current_state,
current_state_root,
&stray_slots,
&adversarial_validators,
)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Precondition: Ensure all stray_blocks blocks are still known
for &block_hash in stray_blocks.values() {
assert!(
rig.block_exists(block_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray block {} should be still present",
block_hash
);
}
for (&slot, &state_hash) in &stray_states {
assert!(
rig.hot_state_exists(state_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray state {} at slot {} should be still present",
state_hash,
slot
);
}
assert_eq!(rig.get_finalized_checkpoints(), hashset! {},);
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
assert!(rig.chain.knows_head(&stray_head));
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Trigger finalization
let finalization_slots: Vec<Slot> = ((canonical_chain_slot + 1)
..=(canonical_chain_slot + slots_per_epoch * 5))
.map(Slot::new)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let state_root = state.update_tree_hash_cache().unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (canonical_chain_blocks_post_finalization, _, _, _) = rig
.add_attested_blocks_at_slots(state, state_root, &finalization_slots, &honest_validators)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Postcondition: New blocks got finalized
assert_eq!(
rig.get_finalized_checkpoints(),
hashset! {
canonical_chain_blocks_pre_finalization[&rig.epoch_start_slot(1).into()],
canonical_chain_blocks_post_finalization[&rig.epoch_start_slot(2).into()],
},
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
);
// Postcondition: Ensure all stray_blocks blocks have been pruned
for &block_hash in stray_blocks.values() {
assert!(
!rig.block_exists(block_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"abandoned block {} should have been pruned",
block_hash
);
}
for (&slot, &state_hash) in &stray_states {
assert!(
!rig.hot_state_exists(state_hash),
"stray state {} at slot {} should have been pruned",
state_hash,
slot
);
assert!(
!rig.cold_state_exists(state_hash),
"stray state {} at slot {} should have been pruned",
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
state_hash,
slot
);
}
assert!(!rig.chain.knows_head(&stray_head));
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn pruning_does_not_touch_abandoned_block_shared_with_canonical_chain() {
const HONEST_VALIDATOR_COUNT: usize = 16 + 0;
const ADVERSARIAL_VALIDATOR_COUNT: usize = 8 - 0;
const VALIDATOR_COUNT: usize = HONEST_VALIDATOR_COUNT + ADVERSARIAL_VALIDATOR_COUNT;
let validators_keypairs = types::test_utils::generate_deterministic_keypairs(VALIDATOR_COUNT);
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
let honest_validators: Vec<usize> = (0..HONEST_VALIDATOR_COUNT).collect();
let adversarial_validators: Vec<usize> = (HONEST_VALIDATOR_COUNT..VALIDATOR_COUNT).collect();
let rig = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.keypairs(validators_keypairs)
.fresh_ephemeral_store()
.mock_execution_layer()
.build();
let slots_per_epoch = rig.slots_per_epoch();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (state, state_root) = rig.get_current_state_and_root();
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Fill up 0th epoch
let canonical_chain_slots_zeroth_epoch: Vec<Slot> =
(1..rig.epoch_start_slot(1)).map(Slot::new).collect();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (_, _, _, mut state) = rig
.add_attested_blocks_at_slots(
state,
state_root,
&canonical_chain_slots_zeroth_epoch,
&honest_validators,
)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Fill up 1st epoch
let canonical_chain_slots_first_epoch: Vec<Slot> = (rig.epoch_start_slot(1)
..=rig.epoch_start_slot(1) + 1)
.map(Slot::new)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let state_root = state.update_tree_hash_cache().unwrap();
let (canonical_chain_blocks_first_epoch, _, shared_head, mut state) = rig
.add_attested_blocks_at_slots(
state.clone(),
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
state_root,
&canonical_chain_slots_first_epoch,
&honest_validators,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
let canonical_chain_slot: u64 = rig.get_current_slot().into();
let stray_chain_slots_first_epoch: Vec<Slot> = (rig.epoch_start_slot(1) + 2
..=rig.epoch_start_slot(1) + 2)
.map(Slot::new)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let state_root = state.update_tree_hash_cache().unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (stray_blocks, stray_states, stray_head, _) = rig
.add_attested_blocks_at_slots(
state.clone(),
state_root,
&stray_chain_slots_first_epoch,
&adversarial_validators,
)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Preconditions
for &block_hash in stray_blocks.values() {
assert!(
rig.block_exists(block_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray block {} should be still present",
block_hash
);
}
for (&slot, &state_hash) in &stray_states {
assert!(
rig.hot_state_exists(state_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray state {} at slot {} should be still present",
state_hash,
slot
);
}
let chain_dump = rig.chain.chain_dump().unwrap();
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
assert_eq!(
get_finalized_epoch_boundary_blocks(&chain_dump),
vec![Hash256::zero().into()].into_iter().collect(),
);
assert!(get_blocks(&chain_dump).contains(&shared_head));
// Trigger finalization
let finalization_slots: Vec<Slot> = ((canonical_chain_slot + 1)
..=(canonical_chain_slot + slots_per_epoch * 5))
.map(Slot::new)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let state_root = state.update_tree_hash_cache().unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (canonical_chain_blocks, _, _, _) = rig
.add_attested_blocks_at_slots(state, state_root, &finalization_slots, &honest_validators)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Postconditions
assert_eq!(
rig.get_finalized_checkpoints(),
hashset! {
canonical_chain_blocks_first_epoch[&rig.epoch_start_slot(1).into()],
canonical_chain_blocks[&rig.epoch_start_slot(2).into()],
},
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
);
for &block_hash in stray_blocks.values() {
assert!(
!rig.block_exists(block_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray block {} should have been pruned",
block_hash,
);
}
for (&slot, &state_hash) in &stray_states {
assert!(
!rig.hot_state_exists(state_hash),
"stray state {} at slot {} should have been pruned",
state_hash,
slot
);
assert!(
!rig.cold_state_exists(state_hash),
"stray state {} at slot {} should have been pruned",
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
state_hash,
slot
);
}
assert!(!rig.chain.knows_head(&stray_head));
let chain_dump = rig.chain.chain_dump().unwrap();
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
assert!(get_blocks(&chain_dump).contains(&shared_head));
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn pruning_does_not_touch_blocks_prior_to_finalization() {
const HONEST_VALIDATOR_COUNT: usize = 16;
const ADVERSARIAL_VALIDATOR_COUNT: usize = 8;
const VALIDATOR_COUNT: usize = HONEST_VALIDATOR_COUNT + ADVERSARIAL_VALIDATOR_COUNT;
let validators_keypairs = types::test_utils::generate_deterministic_keypairs(VALIDATOR_COUNT);
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
let honest_validators: Vec<usize> = (0..HONEST_VALIDATOR_COUNT).collect();
let adversarial_validators: Vec<usize> = (HONEST_VALIDATOR_COUNT..VALIDATOR_COUNT).collect();
let rig = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.keypairs(validators_keypairs)
.fresh_ephemeral_store()
.mock_execution_layer()
.build();
let slots_per_epoch = rig.slots_per_epoch();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (mut state, state_root) = rig.get_current_state_and_root();
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Fill up 0th epoch with canonical chain blocks
let zeroth_epoch_slots: Vec<Slot> = (1..=rig.epoch_start_slot(1)).map(Slot::new).collect();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (canonical_chain_blocks, _, _, new_state) = rig
.add_attested_blocks_at_slots(state, state_root, &zeroth_epoch_slots, &honest_validators)
.await;
state = new_state;
let canonical_chain_slot: u64 = rig.get_current_slot().into();
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Fill up 1st epoch. Contains a fork.
let first_epoch_slots: Vec<Slot> = ((rig.epoch_start_slot(1) + 1)..(rig.epoch_start_slot(2)))
.map(Slot::new)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let state_root = state.update_tree_hash_cache().unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (stray_blocks, stray_states, stray_head, _) = rig
.add_attested_blocks_at_slots(
state.clone(),
state_root,
&first_epoch_slots,
&adversarial_validators,
)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Preconditions
for &block_hash in stray_blocks.values() {
assert!(
rig.block_exists(block_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray block {} should be still present",
block_hash
);
}
for (&slot, &state_hash) in &stray_states {
assert!(
rig.hot_state_exists(state_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray state {} at slot {} should be still present",
state_hash,
slot
);
}
assert_eq!(rig.get_finalized_checkpoints(), hashset! {});
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Trigger finalization
let slots: Vec<Slot> = ((canonical_chain_slot + 1)
..=(canonical_chain_slot + slots_per_epoch * 4))
.map(Slot::new)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let state_root = state.update_tree_hash_cache().unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (_, _, _, _) = rig
.add_attested_blocks_at_slots(state, state_root, &slots, &honest_validators)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Postconditions
assert_eq!(
rig.get_finalized_checkpoints(),
hashset! {canonical_chain_blocks[&rig.epoch_start_slot(1).into()]},
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
);
for &block_hash in stray_blocks.values() {
assert!(
rig.block_exists(block_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray block {} should be still present",
block_hash
);
}
for (&slot, &state_hash) in &stray_states {
assert!(
rig.hot_state_exists(state_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray state {} at slot {} should be still present",
state_hash,
slot
);
}
assert!(rig.chain.knows_head(&stray_head));
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn prunes_fork_growing_past_youngest_finalized_checkpoint() {
const HONEST_VALIDATOR_COUNT: usize = 16 + 0;
const ADVERSARIAL_VALIDATOR_COUNT: usize = 8 - 0;
const VALIDATOR_COUNT: usize = HONEST_VALIDATOR_COUNT + ADVERSARIAL_VALIDATOR_COUNT;
let validators_keypairs = types::test_utils::generate_deterministic_keypairs(VALIDATOR_COUNT);
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
let honest_validators: Vec<usize> = (0..HONEST_VALIDATOR_COUNT).collect();
let adversarial_validators: Vec<usize> = (HONEST_VALIDATOR_COUNT..VALIDATOR_COUNT).collect();
let rig = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.keypairs(validators_keypairs)
.fresh_ephemeral_store()
.mock_execution_layer()
.build();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (state, state_root) = rig.get_current_state_and_root();
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Fill up 0th epoch with canonical chain blocks
let zeroth_epoch_slots: Vec<Slot> = (1..=rig.epoch_start_slot(1)).map(Slot::new).collect();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (canonical_blocks_zeroth_epoch, _, _, mut state) = rig
.add_attested_blocks_at_slots(state, state_root, &zeroth_epoch_slots, &honest_validators)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Fill up 1st epoch. Contains a fork.
let slots_first_epoch: Vec<Slot> = (rig.epoch_start_slot(1) + 1..rig.epoch_start_slot(2))
.map(Into::into)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let state_root = state.update_tree_hash_cache().unwrap();
let (stray_blocks_first_epoch, stray_states_first_epoch, _, mut stray_state) = rig
.add_attested_blocks_at_slots(
state.clone(),
state_root,
&slots_first_epoch,
&adversarial_validators,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
let (canonical_blocks_first_epoch, _, _, mut canonical_state) = rig
.add_attested_blocks_at_slots(state, state_root, &slots_first_epoch, &honest_validators)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Fill up 2nd epoch. Extends both the canonical chain and the fork.
let stray_slots_second_epoch: Vec<Slot> = (rig.epoch_start_slot(2)
..=rig.epoch_start_slot(2) + 1)
.map(Into::into)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let stray_state_root = stray_state.update_tree_hash_cache().unwrap();
let (stray_blocks_second_epoch, stray_states_second_epoch, stray_head, _) = rig
.add_attested_blocks_at_slots(
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
stray_state,
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
stray_state_root,
&stray_slots_second_epoch,
&adversarial_validators,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Precondition: Ensure all stray_blocks blocks are still known
let stray_blocks: HashMap<Slot, SignedBeaconBlockHash> = stray_blocks_first_epoch
.into_iter()
.chain(stray_blocks_second_epoch.into_iter())
.collect();
let stray_states: HashMap<Slot, BeaconStateHash> = stray_states_first_epoch
.into_iter()
.chain(stray_states_second_epoch.into_iter())
.collect();
for &block_hash in stray_blocks.values() {
assert!(
rig.block_exists(block_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray block {} should be still present",
block_hash
);
}
for (&slot, &state_hash) in &stray_states {
assert!(
rig.hot_state_exists(state_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray state {} at slot {} should be still present",
state_hash,
slot
);
}
// Precondition: Nothing is finalized yet
assert_eq!(rig.get_finalized_checkpoints(), hashset! {},);
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
assert!(rig.chain.knows_head(&stray_head));
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Trigger finalization
let canonical_slots: Vec<Slot> = (rig.epoch_start_slot(2)..=rig.epoch_start_slot(6))
.map(Into::into)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let canonical_state_root = canonical_state.update_tree_hash_cache().unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (canonical_blocks, _, _, _) = rig
.add_attested_blocks_at_slots(
canonical_state,
canonical_state_root,
&canonical_slots,
&honest_validators,
)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Postconditions
let canonical_blocks: HashMap<Slot, SignedBeaconBlockHash> = canonical_blocks_zeroth_epoch
.into_iter()
.chain(canonical_blocks_first_epoch.into_iter())
.chain(canonical_blocks.into_iter())
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
.collect();
// Postcondition: New blocks got finalized
assert_eq!(
rig.get_finalized_checkpoints(),
hashset! {
canonical_blocks[&rig.epoch_start_slot(1).into()],
canonical_blocks[&rig.epoch_start_slot(2).into()],
},
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
);
// Postcondition: Ensure all stray_blocks blocks have been pruned
for &block_hash in stray_blocks.values() {
assert!(
!rig.block_exists(block_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"abandoned block {} should have been pruned",
block_hash
);
}
for (&slot, &state_hash) in &stray_states {
assert!(
!rig.hot_state_exists(state_hash),
"stray state {} at slot {} should have been pruned",
state_hash,
slot
);
assert!(
!rig.cold_state_exists(state_hash),
"stray state {} at slot {} should have been pruned",
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
state_hash,
slot
);
}
assert!(!rig.chain.knows_head(&stray_head));
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
}
// This is to check if state outside of normal block processing are pruned correctly.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn prunes_skipped_slots_states() {
const HONEST_VALIDATOR_COUNT: usize = 16 + 0;
const ADVERSARIAL_VALIDATOR_COUNT: usize = 8 - 0;
const VALIDATOR_COUNT: usize = HONEST_VALIDATOR_COUNT + ADVERSARIAL_VALIDATOR_COUNT;
let validators_keypairs = types::test_utils::generate_deterministic_keypairs(VALIDATOR_COUNT);
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
let honest_validators: Vec<usize> = (0..HONEST_VALIDATOR_COUNT).collect();
let adversarial_validators: Vec<usize> = (HONEST_VALIDATOR_COUNT..VALIDATOR_COUNT).collect();
let rig = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.keypairs(validators_keypairs)
.fresh_ephemeral_store()
.mock_execution_layer()
.build();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (state, state_root) = rig.get_current_state_and_root();
let canonical_slots_zeroth_epoch: Vec<Slot> =
(1..=rig.epoch_start_slot(1)).map(Into::into).collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (canonical_blocks_zeroth_epoch, _, _, mut canonical_state) = rig
.add_attested_blocks_at_slots(
state.clone(),
state_root,
&canonical_slots_zeroth_epoch,
&honest_validators,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
let skipped_slot: Slot = (rig.epoch_start_slot(1) + 1).into();
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
let stray_slots: Vec<Slot> = ((skipped_slot + 1).into()..rig.epoch_start_slot(2))
.map(Into::into)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let canonical_state_root = canonical_state.update_tree_hash_cache().unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (stray_blocks, stray_states, _, stray_state) = rig
.add_attested_blocks_at_slots(
canonical_state.clone(),
canonical_state_root,
&stray_slots,
&adversarial_validators,
)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Preconditions
for &block_hash in stray_blocks.values() {
assert!(
rig.block_exists(block_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray block {} should be still present",
block_hash
);
}
for (&slot, &state_hash) in &stray_states {
assert!(
rig.hot_state_exists(state_hash),
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
"stray state {} at slot {} should be still present",
state_hash,
slot
);
}
assert_eq!(rig.get_finalized_checkpoints(), hashset! {},);
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Make sure slots were skipped
assert!(rig.is_skipped_slot(&stray_state, skipped_slot));
{
let state_hash = (*stray_state.get_state_root(skipped_slot).unwrap()).into();
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
assert!(
rig.hot_state_exists(state_hash),
"skipped slot state {} should be still present",
state_hash
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
);
}
// Trigger finalization
let canonical_slots: Vec<Slot> = ((skipped_slot + 1).into()..rig.epoch_start_slot(7))
.map(Into::into)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let canonical_state_root = canonical_state.update_tree_hash_cache().unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (canonical_blocks_post_finalization, _, _, _) = rig
.add_attested_blocks_at_slots(
canonical_state,
canonical_state_root,
&canonical_slots,
&honest_validators,
)
.await;
// Postconditions
let canonical_blocks: HashMap<Slot, SignedBeaconBlockHash> = canonical_blocks_zeroth_epoch
.into_iter()
.chain(canonical_blocks_post_finalization.into_iter())
.collect();
assert_eq!(
rig.get_finalized_checkpoints(),
hashset! {
canonical_blocks[&rig.epoch_start_slot(1).into()],
canonical_blocks[&rig.epoch_start_slot(2).into()],
},
);
for (&slot, &state_hash) in &stray_states {
assert!(
!rig.hot_state_exists(state_hash),
"stray state {} at slot {} should have been pruned",
state_hash,
slot
);
assert!(
!rig.cold_state_exists(state_hash),
"stray state {} at slot {} should have been pruned",
state_hash,
slot
);
}
assert!(rig.is_skipped_slot(&stray_state, skipped_slot));
{
let state_hash: BeaconStateHash =
(*stray_state.get_state_root(skipped_slot).unwrap()).into();
assert!(
!rig.hot_state_exists(state_hash),
"skipped slot {} state {} should have been pruned",
skipped_slot,
state_hash
);
}
}
// This is to check if state outside of normal block processing are pruned correctly.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn finalizes_non_epoch_start_slot() {
const HONEST_VALIDATOR_COUNT: usize = 16 + 0;
const ADVERSARIAL_VALIDATOR_COUNT: usize = 8 - 0;
const VALIDATOR_COUNT: usize = HONEST_VALIDATOR_COUNT + ADVERSARIAL_VALIDATOR_COUNT;
let validators_keypairs = types::test_utils::generate_deterministic_keypairs(VALIDATOR_COUNT);
let honest_validators: Vec<usize> = (0..HONEST_VALIDATOR_COUNT).collect();
let adversarial_validators: Vec<usize> = (HONEST_VALIDATOR_COUNT..VALIDATOR_COUNT).collect();
let rig = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.keypairs(validators_keypairs)
.fresh_ephemeral_store()
.mock_execution_layer()
.build();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (state, state_root) = rig.get_current_state_and_root();
let canonical_slots_zeroth_epoch: Vec<Slot> =
(1..rig.epoch_start_slot(1)).map(Into::into).collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (canonical_blocks_zeroth_epoch, _, _, mut canonical_state) = rig
.add_attested_blocks_at_slots(
state.clone(),
state_root,
&canonical_slots_zeroth_epoch,
&honest_validators,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
let skipped_slot: Slot = rig.epoch_start_slot(1).into();
let stray_slots: Vec<Slot> = ((skipped_slot + 1).into()..rig.epoch_start_slot(2))
.map(Into::into)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let canonical_state_root = canonical_state.update_tree_hash_cache().unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (stray_blocks, stray_states, _, stray_state) = rig
.add_attested_blocks_at_slots(
canonical_state.clone(),
canonical_state_root,
&stray_slots,
&adversarial_validators,
)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Preconditions
for &block_hash in stray_blocks.values() {
assert!(
rig.block_exists(block_hash),
"stray block {} should be still present",
block_hash
);
}
for (&slot, &state_hash) in &stray_states {
assert!(
rig.hot_state_exists(state_hash),
"stray state {} at slot {} should be still present",
state_hash,
slot
);
}
assert_eq!(rig.get_finalized_checkpoints(), hashset! {});
// Make sure slots were skipped
assert!(rig.is_skipped_slot(&stray_state, skipped_slot));
{
let state_hash = (*stray_state.get_state_root(skipped_slot).unwrap()).into();
assert!(
rig.hot_state_exists(state_hash),
"skipped slot state {} should be still present",
state_hash
);
}
// Trigger finalization
let canonical_slots: Vec<Slot> = ((skipped_slot + 1).into()..rig.epoch_start_slot(7))
.map(Into::into)
.collect();
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let canonical_state_root = canonical_state.update_tree_hash_cache().unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (canonical_blocks_post_finalization, _, _, _) = rig
.add_attested_blocks_at_slots(
canonical_state,
canonical_state_root,
&canonical_slots,
&honest_validators,
)
.await;
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
// Postconditions
let canonical_blocks: HashMap<Slot, SignedBeaconBlockHash> = canonical_blocks_zeroth_epoch
.into_iter()
.chain(canonical_blocks_post_finalization.into_iter())
.collect();
assert_eq!(
rig.get_finalized_checkpoints(),
hashset! {
canonical_blocks[&(rig.epoch_start_slot(1)-1).into()],
canonical_blocks[&rig.epoch_start_slot(2).into()],
},
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
);
for (&slot, &state_hash) in &stray_states {
assert!(
!rig.hot_state_exists(state_hash),
"stray state {} at slot {} should have been pruned",
state_hash,
slot
);
assert!(
!rig.cold_state_exists(state_hash),
"stray state {} at slot {} should have been pruned",
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
state_hash,
slot
);
}
assert!(rig.is_skipped_slot(&stray_state, skipped_slot));
{
let state_hash: BeaconStateHash =
(*stray_state.get_state_root(skipped_slot).unwrap()).into();
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
assert!(
!rig.hot_state_exists(state_hash),
Add attestation gossip pre-verification (#983) * Add PH & MS slot clock changes * Account for genesis time * Add progress on duties refactor * Add simple is_aggregator bool to val subscription * Start work on attestation_verification.rs * Add progress on ObservedAttestations * Progress with ObservedAttestations * Fix tests * Add observed attestations to the beacon chain * Add attestation observation to processing code * Add progress on attestation verification * Add first draft of ObservedAttesters * Add more tests * Add observed attesters to beacon chain * Add observers to attestation processing * Add more attestation verification * Create ObservedAggregators map * Remove commented-out code * Add observed aggregators into chain * Add progress * Finish adding features to attestation verification * Ensure beacon chain compiles * Link attn verification into chain * Integrate new attn verification in chain * Remove old attestation processing code * Start trying to fix beacon_chain tests * Split adding into pools into two functions * Add aggregation to harness * Get test harness working again * Adjust the number of aggregators for test harness * Fix edge-case in harness * Integrate new attn processing in network * Fix compile bug in validator_client * Update validator API endpoints * Fix aggreagation in test harness * Fix enum thing * Fix attestation observation bug: * Patch failing API tests * Start adding comments to attestation verification * Remove unused attestation field * Unify "is block known" logic * Update comments * Supress fork choice errors for network processing * Add todos * Tidy * Add gossip attn tests * Disallow test harness to produce old attns * Comment out in-progress tests * Partially address pruning tests * Fix failing store test * Add aggregate tests * Add comments about which spec conditions we check * Dont re-aggregate * Split apart test harness attn production * Fix compile error in network * Make progress on commented-out test * Fix skipping attestation test * Add fork choice verification tests * Tidy attn tests, remove dead code * Remove some accidentally added code * Fix clippy lint * Rename test file * Add block tests, add cheap block proposer check * Rename block testing file * Add observed_block_producers * Tidy * Switch around block signature verification * Finish block testing * Remove gossip from signature tests * First pass of self review * Fix deviation in spec * Update test spec tags * Start moving over to hashset * Finish moving observed attesters to hashmap * Move aggregation pool over to hashmap * Make fc attn borrow again * Fix rest_api compile error * Fix missing comments * Fix monster test * Uncomment increasing slots test * Address remaining comments * Remove unsafe, use cfg test * Remove cfg test flag * Fix dodgy comment * Ignore aggregates that are already known. * Unify aggregator modulo logic * Fix typo in logs * Refactor validator subscription logic * Avoid reproducing selection proof * Skip HTTP call if no subscriptions * Rename DutyAndState -> DutyAndProof * Tidy logs * Print root as dbg * Fix compile errors in tests * Fix compile error in test
2020-05-06 11:42:56 +00:00
"skipped slot {} state {} should have been pruned",
skipped_slot,
Add attestation gossip pre-verification (#983) * Add PH & MS slot clock changes * Account for genesis time * Add progress on duties refactor * Add simple is_aggregator bool to val subscription * Start work on attestation_verification.rs * Add progress on ObservedAttestations * Progress with ObservedAttestations * Fix tests * Add observed attestations to the beacon chain * Add attestation observation to processing code * Add progress on attestation verification * Add first draft of ObservedAttesters * Add more tests * Add observed attesters to beacon chain * Add observers to attestation processing * Add more attestation verification * Create ObservedAggregators map * Remove commented-out code * Add observed aggregators into chain * Add progress * Finish adding features to attestation verification * Ensure beacon chain compiles * Link attn verification into chain * Integrate new attn verification in chain * Remove old attestation processing code * Start trying to fix beacon_chain tests * Split adding into pools into two functions * Add aggregation to harness * Get test harness working again * Adjust the number of aggregators for test harness * Fix edge-case in harness * Integrate new attn processing in network * Fix compile bug in validator_client * Update validator API endpoints * Fix aggreagation in test harness * Fix enum thing * Fix attestation observation bug: * Patch failing API tests * Start adding comments to attestation verification * Remove unused attestation field * Unify "is block known" logic * Update comments * Supress fork choice errors for network processing * Add todos * Tidy * Add gossip attn tests * Disallow test harness to produce old attns * Comment out in-progress tests * Partially address pruning tests * Fix failing store test * Add aggregate tests * Add comments about which spec conditions we check * Dont re-aggregate * Split apart test harness attn production * Fix compile error in network * Make progress on commented-out test * Fix skipping attestation test * Add fork choice verification tests * Tidy attn tests, remove dead code * Remove some accidentally added code * Fix clippy lint * Rename test file * Add block tests, add cheap block proposer check * Rename block testing file * Add observed_block_producers * Tidy * Switch around block signature verification * Finish block testing * Remove gossip from signature tests * First pass of self review * Fix deviation in spec * Update test spec tags * Start moving over to hashset * Finish moving observed attesters to hashmap * Move aggregation pool over to hashmap * Make fc attn borrow again * Fix rest_api compile error * Fix missing comments * Fix monster test * Uncomment increasing slots test * Address remaining comments * Remove unsafe, use cfg test * Remove cfg test flag * Fix dodgy comment * Ignore aggregates that are already known. * Unify aggregator modulo logic * Fix typo in logs * Refactor validator subscription logic * Avoid reproducing selection proof * Skip HTTP call if no subscriptions * Rename DutyAndState -> DutyAndProof * Tidy logs * Print root as dbg * Fix compile errors in tests * Fix compile error in test
2020-05-06 11:42:56 +00:00
state_hash
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
);
}
}
fn check_all_blocks_exist<'a>(
harness: &TestHarness,
blocks: impl Iterator<Item = &'a SignedBeaconBlockHash>,
) {
for &block_hash in blocks {
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
let block = harness.chain.get_blinded_block(&block_hash.into()).unwrap();
assert!(
block.is_some(),
"expected block {:?} to be in DB",
block_hash
);
}
}
fn check_all_states_exist<'a>(
harness: &TestHarness,
states: impl Iterator<Item = &'a BeaconStateHash>,
) {
for &state_hash in states {
let state = harness.chain.get_state(&state_hash.into(), None).unwrap();
assert!(
state.is_some(),
"expected state {:?} to be in DB",
state_hash,
);
}
}
// Check that none of the given states exist in the database.
fn check_no_states_exist<'a>(
harness: &TestHarness,
states: impl Iterator<Item = &'a BeaconStateHash>,
) {
for &state_root in states {
assert!(
harness
.chain
.get_state(&state_root.into(), None)
.unwrap()
.is_none(),
"state {:?} should not be in the DB",
state_root
);
}
}
// Check that none of the given blocks exist in the database.
fn check_no_blocks_exist<'a>(
harness: &TestHarness,
blocks: impl Iterator<Item = &'a SignedBeaconBlockHash>,
) {
for &block_hash in blocks {
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
let block = harness.chain.get_blinded_block(&block_hash.into()).unwrap();
assert!(
block.is_none(),
"did not expect block {:?} to be in the DB",
block_hash
);
}
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn prune_single_block_fork() {
let slots_per_epoch = E::slots_per_epoch();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
pruning_test(3 * slots_per_epoch, 1, slots_per_epoch, 0, 1).await;
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn prune_single_block_long_skip() {
let slots_per_epoch = E::slots_per_epoch();
pruning_test(
2 * slots_per_epoch,
1,
2 * slots_per_epoch,
2 * slots_per_epoch as u64,
1,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn prune_shared_skip_states_mid_epoch() {
let slots_per_epoch = E::slots_per_epoch();
pruning_test(
slots_per_epoch + slots_per_epoch / 2,
1,
slots_per_epoch,
2,
slots_per_epoch - 1,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn prune_shared_skip_states_epoch_boundaries() {
let slots_per_epoch = E::slots_per_epoch();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
pruning_test(slots_per_epoch - 1, 1, slots_per_epoch, 2, slots_per_epoch).await;
pruning_test(slots_per_epoch - 1, 2, slots_per_epoch, 1, slots_per_epoch).await;
pruning_test(
2 * slots_per_epoch + slots_per_epoch / 2,
slots_per_epoch as u64 / 2,
slots_per_epoch,
slots_per_epoch as u64 / 2 + 1,
slots_per_epoch,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
pruning_test(
2 * slots_per_epoch + slots_per_epoch / 2,
slots_per_epoch as u64 / 2,
slots_per_epoch,
slots_per_epoch as u64 / 2 + 1,
slots_per_epoch,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
pruning_test(
2 * slots_per_epoch - 1,
slots_per_epoch as u64,
1,
0,
2 * slots_per_epoch,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
)
.await;
}
/// Generic harness for pruning tests.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
async fn pruning_test(
// Number of blocks to start the chain with before forking.
num_initial_blocks: u64,
// Number of skip slots on the main chain after the initial blocks.
num_canonical_skips: u64,
// Number of blocks on the main chain after the skip, but before the finalisation-triggering
// blocks.
num_canonical_middle_blocks: u64,
// Number of skip slots on the fork chain after the initial blocks.
num_fork_skips: u64,
// Number of blocks on the fork chain after the skips.
num_fork_blocks: u64,
) {
const VALIDATOR_COUNT: usize = 24;
const VALIDATOR_SUPERMAJORITY: usize = (VALIDATOR_COUNT / 3) * 2;
const HONEST_VALIDATOR_COUNT: usize = VALIDATOR_SUPERMAJORITY;
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
Fix head tracker concurrency bugs (#1771) ## Issue Addressed Closes #1557 ## Proposed Changes Modify the pruning algorithm so that it mutates the head-tracker _before_ committing the database transaction to disk, and _only if_ all the heads to be removed are still present in the head-tracker (i.e. no concurrent mutations). In the process of writing and testing this I also had to make a few other changes: * Use internal mutability for all `BeaconChainHarness` functions (namely the RNG and the graffiti), in order to enable parallel calls (see testing section below). * Disable logging in harness tests unless the `test_logger` feature is turned on And chose to make some clean-ups: * Delete the `NullMigrator` * Remove type-based configuration for the migrator in favour of runtime config (simpler, less duplicated code) * Use the non-blocking migrator unless the blocking migrator is required. In the store tests we need the blocking migrator because some tests make asserts about the state of the DB after the migration has run. * Rename `validators_keypairs` -> `validator_keypairs` in the `BeaconChainHarness` ## Testing To confirm that the fix worked, I wrote a test using [Hiatus](https://crates.io/crates/hiatus), which can be found here: https://github.com/michaelsproul/lighthouse/tree/hiatus-issue-1557 That test can't be merged because it inserts random breakpoints everywhere, but if you check out that branch you can run the test with: ``` $ cd beacon_node/beacon_chain $ cargo test --release --test parallel_tests --features test_logger ``` It should pass, and the log output should show: ``` WARN Pruning deferred because of a concurrent mutation, message: this is expected only very rarely! ``` ## Additional Info This is a backwards-compatible change with no impact on consensus.
2020-10-19 05:58:39 +00:00
let harness = get_harness(store.clone(), VALIDATOR_COUNT);
let honest_validators: Vec<usize> = (0..HONEST_VALIDATOR_COUNT).collect();
let faulty_validators: Vec<usize> = (HONEST_VALIDATOR_COUNT..VALIDATOR_COUNT).collect();
let slots = |start: Slot, num_blocks: u64| -> Vec<Slot> {
(start.as_u64()..start.as_u64() + num_blocks)
.map(Slot::new)
.collect()
};
let start_slot = Slot::new(1);
let divergence_slot = start_slot + num_initial_blocks;
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (state, state_root) = harness.get_current_state_and_root();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (_, _, _, divergence_state) = harness
.add_attested_blocks_at_slots(
state,
state_root,
&slots(start_slot, num_initial_blocks)[..],
&honest_validators,
)
.await;
let mut chains = harness
.add_blocks_on_multiple_chains(vec![
// Canonical chain
(
divergence_state.clone(),
slots(
divergence_slot + num_canonical_skips,
num_canonical_middle_blocks,
),
honest_validators.clone(),
),
// Fork chain
(
divergence_state.clone(),
slots(divergence_slot + num_fork_skips, num_fork_blocks),
faulty_validators,
),
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
])
.await;
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let (_, _, _, mut canonical_state) = chains.remove(0);
let (stray_blocks, stray_states, _, stray_head_state) = chains.remove(0);
let stray_head_slot = divergence_slot + num_fork_skips + num_fork_blocks - 1;
let stray_head_state_root = stray_states[&stray_head_slot];
let stray_states = harness
.chain
.rev_iter_state_roots_from(stray_head_state_root.into(), &stray_head_state)
.map(Result::unwrap)
.map(|(state_root, _)| state_root.into())
.collect::<HashSet<_>>();
check_all_blocks_exist(&harness, stray_blocks.values());
check_all_states_exist(&harness, stray_states.iter());
let chain_dump = harness.chain.chain_dump().unwrap();
assert_eq!(
get_finalized_epoch_boundary_blocks(&chain_dump),
vec![Hash256::zero().into()].into_iter().collect(),
);
// Trigger finalization
let num_finalization_blocks = 4 * E::slots_per_epoch();
let canonical_slot = divergence_slot + num_canonical_skips + num_canonical_middle_blocks;
Optimize validator duties (#2243) ## Issue Addressed Closes #2052 ## Proposed Changes - Refactor the attester/proposer duties endpoints in the BN - Performance improvements - Fixes some potential inconsistencies with the dependent root fields. - Removes `http_api::beacon_proposer_cache` and just uses the one on the `BeaconChain` instead. - Move the code for the proposer/attester duties endpoints into separate files, for readability. - Refactor the `DutiesService` in the VC - Required to reduce the delay on broadcasting new blocks. - Gets rid of the `ValidatorDuty` shim struct that came about when we adopted the standard API. - Separate block/attestation duty tasks so that they don't block each other when one is slow. - In the VC, use `PublicKeyBytes` to represent validators instead of `PublicKey`. `PublicKey` is a legit crypto object whilst `PublicKeyBytes` is just a byte-array, it's much faster to clone/hash `PublicKeyBytes` and this change has had a significant impact on runtimes. - Unfortunately this has created lots of dust changes. - In the BN, store `PublicKeyBytes` in the `beacon_proposer_cache` and allow access to them. The HTTP API always sends `PublicKeyBytes` over the wire and the conversion from `PublicKey` -> `PublickeyBytes` is non-trivial, especially when queries have 100s/1000s of validators (like Pyrmont). - Add the `state_processing::state_advance` mod which dedups a lot of the "apply `n` skip slots to the state" code. - This also fixes a bug with some functions which were failing to include a state root as per [this comment](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/consensus/state_processing/src/state_advance.rs#L69-L74). I couldn't find any instance of this bug that resulted in anything more severe than keying a shuffling cache by the wrong block root. - Swap the VC block service to use `mpsc` from `tokio` instead of `futures`. This is consistent with the rest of the code base. ~~This PR *reduces* the size of the codebase :tada:~~ It *used* to reduce the size of the code base before I added more comments. ## Observations on Prymont - Proposer duties times down from peaks of 450ms to consistent <1ms. - Current epoch attester duties times down from >1s peaks to a consistent 20-30ms. - Block production down from +600ms to 100-200ms. ## Additional Info - ~~Blocked on #2241~~ - ~~Blocked on #2234~~ ## TODO - [x] ~~Refactor this into some smaller PRs?~~ Leaving this as-is for now. - [x] Address `per_slot_processing` roots. - [x] Investigate slow next epoch times. Not getting added to cache on block processing? - [x] Consider [this](https://github.com/sigp/lighthouse/blob/072695284f7eff82c51f79bc921ad942fea7483a/beacon_node/store/src/hot_cold_store.rs#L811-L812) in the scenario of replacing the state roots Co-authored-by: pawan <pawandhananjay@gmail.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2021-03-17 05:09:57 +00:00
let canonical_state_root = canonical_state.update_tree_hash_cache().unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.add_attested_blocks_at_slots(
canonical_state,
canonical_state_root,
&slots(canonical_slot, num_finalization_blocks),
&honest_validators,
)
.await;
// Check that finalization has advanced past the divergence slot.
assert!(
harness
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
.finalized_checkpoint()
.epoch
.start_slot(E::slots_per_epoch())
> divergence_slot
);
check_chain_dump(
&harness,
(num_initial_blocks + num_canonical_middle_blocks + num_finalization_blocks + 1) as u64,
);
let all_canonical_states = harness
.chain
.forwards_iter_state_roots(Slot::new(0))
.unwrap()
.map(Result::unwrap)
.map(|(state_root, _)| state_root.into())
.collect::<HashSet<BeaconStateHash>>();
check_all_states_exist(&harness, all_canonical_states.iter());
check_no_states_exist(&harness, stray_states.difference(&all_canonical_states));
check_no_blocks_exist(&harness, stray_blocks.values());
}
#[tokio::test]
async fn garbage_collect_temp_states_from_failed_block() {
Implement database temp states to reduce memory usage (#1798) ## Issue Addressed Closes #800 Closes #1713 ## Proposed Changes Implement the temporary state storage algorithm described in #800. Specifically: * Add `DBColumn::BeaconStateTemporary`, for storing 0-length temporary marker values. * Store intermediate states immediately as they are created, marked temporary. Delete the temporary flag if the block is processed successfully. * Add a garbage collection process to delete leftover temporary states on start-up. * Bump the database schema version to 2 so that a DB with temporary states can't accidentally be used with older versions of the software. The auto-migration is a no-op, but puts in place some infra that we can use for future migrations (e.g. #1784) ## Additional Info There are two known race conditions, one potentially causing permanent faults (hopefully rare), and the other insignificant. ### Race 1: Permanent state marked temporary EDIT: this has been fixed by the addition of a lock around the relevant critical section There are 2 threads that are trying to store 2 different blocks that share some intermediate states (e.g. they both skip some slots from the current head). Consider this sequence of events: 1. Thread 1 checks if state `s` already exists, and seeing that it doesn't, prepares an atomic commit of `(s, s_temporary_flag)`. 2. Thread 2 does the same, but also gets as far as committing the state txn, finishing the processing of its block, and _deleting_ the temporary flag. 3. Thread 1 is (finally) scheduled again, and marks `s` as temporary with its transaction. 4. a) The process is killed, or thread 1's block fails verification and the temp flag is not deleted. This is a permanent failure! Any attempt to load state `s` will fail... hope it isn't on the main chain! Alternatively (4b) happens... b) Thread 1 finishes, and re-deletes the temporary flag. In this case the failure is transient, state `s` will disappear temporarily, but will come back once thread 1 finishes running. I _hope_ that steps 1-3 only happen very rarely, and 4a even more rarely. It's hard to know This once again begs the question of why we're using LevelDB (#483), when it clearly doesn't care about atomicity! A ham-fisted fix would be to wrap the hot and cold DBs in locks, which would bring us closer to how other DBs handle read-write transactions. E.g. [LMDB only allows one R/W transaction at a time](https://docs.rs/lmdb/0.8.0/lmdb/struct.Environment.html#method.begin_rw_txn). ### Race 2: Temporary state returned from `get_state` I don't think this race really matters, but in `load_hot_state`, if another thread stores a state between when we call `load_state_temporary_flag` and when we call `load_hot_state_summary`, then we could end up returning that state even though it's only a temporary state. I can't think of any case where this would be relevant, and I suspect if it did come up, it would be safe/recoverable (having data is safer than _not_ having data). This could be fixed by using a LevelDB read snapshot, but that would require substantial changes to how we read all our values, so I don't think it's worth it right now.
2020-10-23 01:27:51 +00:00
let db_path = tempdir().unwrap();
// Wrap these functions to ensure the variables are dropped before we try to open another
// instance of the store.
let mut store = {
let store = get_store(&db_path);
let harness = get_harness(store.clone(), LOW_VALIDATOR_COUNT);
Implement database temp states to reduce memory usage (#1798) ## Issue Addressed Closes #800 Closes #1713 ## Proposed Changes Implement the temporary state storage algorithm described in #800. Specifically: * Add `DBColumn::BeaconStateTemporary`, for storing 0-length temporary marker values. * Store intermediate states immediately as they are created, marked temporary. Delete the temporary flag if the block is processed successfully. * Add a garbage collection process to delete leftover temporary states on start-up. * Bump the database schema version to 2 so that a DB with temporary states can't accidentally be used with older versions of the software. The auto-migration is a no-op, but puts in place some infra that we can use for future migrations (e.g. #1784) ## Additional Info There are two known race conditions, one potentially causing permanent faults (hopefully rare), and the other insignificant. ### Race 1: Permanent state marked temporary EDIT: this has been fixed by the addition of a lock around the relevant critical section There are 2 threads that are trying to store 2 different blocks that share some intermediate states (e.g. they both skip some slots from the current head). Consider this sequence of events: 1. Thread 1 checks if state `s` already exists, and seeing that it doesn't, prepares an atomic commit of `(s, s_temporary_flag)`. 2. Thread 2 does the same, but also gets as far as committing the state txn, finishing the processing of its block, and _deleting_ the temporary flag. 3. Thread 1 is (finally) scheduled again, and marks `s` as temporary with its transaction. 4. a) The process is killed, or thread 1's block fails verification and the temp flag is not deleted. This is a permanent failure! Any attempt to load state `s` will fail... hope it isn't on the main chain! Alternatively (4b) happens... b) Thread 1 finishes, and re-deletes the temporary flag. In this case the failure is transient, state `s` will disappear temporarily, but will come back once thread 1 finishes running. I _hope_ that steps 1-3 only happen very rarely, and 4a even more rarely. It's hard to know This once again begs the question of why we're using LevelDB (#483), when it clearly doesn't care about atomicity! A ham-fisted fix would be to wrap the hot and cold DBs in locks, which would bring us closer to how other DBs handle read-write transactions. E.g. [LMDB only allows one R/W transaction at a time](https://docs.rs/lmdb/0.8.0/lmdb/struct.Environment.html#method.begin_rw_txn). ### Race 2: Temporary state returned from `get_state` I don't think this race really matters, but in `load_hot_state`, if another thread stores a state between when we call `load_state_temporary_flag` and when we call `load_hot_state_summary`, then we could end up returning that state even though it's only a temporary state. I can't think of any case where this would be relevant, and I suspect if it did come up, it would be safe/recoverable (having data is safer than _not_ having data). This could be fixed by using a LevelDB read snapshot, but that would require substantial changes to how we read all our values, so I don't think it's worth it right now.
2020-10-23 01:27:51 +00:00
let slots_per_epoch = E::slots_per_epoch();
Implement database temp states to reduce memory usage (#1798) ## Issue Addressed Closes #800 Closes #1713 ## Proposed Changes Implement the temporary state storage algorithm described in #800. Specifically: * Add `DBColumn::BeaconStateTemporary`, for storing 0-length temporary marker values. * Store intermediate states immediately as they are created, marked temporary. Delete the temporary flag if the block is processed successfully. * Add a garbage collection process to delete leftover temporary states on start-up. * Bump the database schema version to 2 so that a DB with temporary states can't accidentally be used with older versions of the software. The auto-migration is a no-op, but puts in place some infra that we can use for future migrations (e.g. #1784) ## Additional Info There are two known race conditions, one potentially causing permanent faults (hopefully rare), and the other insignificant. ### Race 1: Permanent state marked temporary EDIT: this has been fixed by the addition of a lock around the relevant critical section There are 2 threads that are trying to store 2 different blocks that share some intermediate states (e.g. they both skip some slots from the current head). Consider this sequence of events: 1. Thread 1 checks if state `s` already exists, and seeing that it doesn't, prepares an atomic commit of `(s, s_temporary_flag)`. 2. Thread 2 does the same, but also gets as far as committing the state txn, finishing the processing of its block, and _deleting_ the temporary flag. 3. Thread 1 is (finally) scheduled again, and marks `s` as temporary with its transaction. 4. a) The process is killed, or thread 1's block fails verification and the temp flag is not deleted. This is a permanent failure! Any attempt to load state `s` will fail... hope it isn't on the main chain! Alternatively (4b) happens... b) Thread 1 finishes, and re-deletes the temporary flag. In this case the failure is transient, state `s` will disappear temporarily, but will come back once thread 1 finishes running. I _hope_ that steps 1-3 only happen very rarely, and 4a even more rarely. It's hard to know This once again begs the question of why we're using LevelDB (#483), when it clearly doesn't care about atomicity! A ham-fisted fix would be to wrap the hot and cold DBs in locks, which would bring us closer to how other DBs handle read-write transactions. E.g. [LMDB only allows one R/W transaction at a time](https://docs.rs/lmdb/0.8.0/lmdb/struct.Environment.html#method.begin_rw_txn). ### Race 2: Temporary state returned from `get_state` I don't think this race really matters, but in `load_hot_state`, if another thread stores a state between when we call `load_state_temporary_flag` and when we call `load_hot_state_summary`, then we could end up returning that state even though it's only a temporary state. I can't think of any case where this would be relevant, and I suspect if it did come up, it would be safe/recoverable (having data is safer than _not_ having data). This could be fixed by using a LevelDB read snapshot, but that would require substantial changes to how we read all our values, so I don't think it's worth it right now.
2020-10-23 01:27:51 +00:00
let genesis_state = harness.get_current_state();
let block_slot = Slot::new(2 * slots_per_epoch);
let (signed_block, state) = harness.make_block(genesis_state, block_slot).await;
let (mut block, _) = signed_block.deconstruct();
// Mutate the block to make it invalid, and re-sign it.
*block.state_root_mut() = Hash256::repeat_byte(0xff);
let proposer_index = block.proposer_index() as usize;
let block = block.sign(
&harness.validator_keypairs[proposer_index].sk,
&state.fork(),
state.genesis_validators_root(),
&harness.spec,
);
// The block should be rejected, but should store a bunch of temporary states.
harness.set_current_slot(block_slot);
harness.process_block_result(block).await.unwrap_err();
assert_eq!(
store.iter_temporary_state_roots().count(),
block_slot.as_usize() - 1
);
store
};
// Wait until all the references to the store have been dropped, this helps ensure we can
// re-open the store later.
loop {
store = if let Err(store_arc) = Arc::try_unwrap(store) {
sleep(Duration::from_millis(500)).await;
store_arc
} else {
break;
}
}
Implement database temp states to reduce memory usage (#1798) ## Issue Addressed Closes #800 Closes #1713 ## Proposed Changes Implement the temporary state storage algorithm described in #800. Specifically: * Add `DBColumn::BeaconStateTemporary`, for storing 0-length temporary marker values. * Store intermediate states immediately as they are created, marked temporary. Delete the temporary flag if the block is processed successfully. * Add a garbage collection process to delete leftover temporary states on start-up. * Bump the database schema version to 2 so that a DB with temporary states can't accidentally be used with older versions of the software. The auto-migration is a no-op, but puts in place some infra that we can use for future migrations (e.g. #1784) ## Additional Info There are two known race conditions, one potentially causing permanent faults (hopefully rare), and the other insignificant. ### Race 1: Permanent state marked temporary EDIT: this has been fixed by the addition of a lock around the relevant critical section There are 2 threads that are trying to store 2 different blocks that share some intermediate states (e.g. they both skip some slots from the current head). Consider this sequence of events: 1. Thread 1 checks if state `s` already exists, and seeing that it doesn't, prepares an atomic commit of `(s, s_temporary_flag)`. 2. Thread 2 does the same, but also gets as far as committing the state txn, finishing the processing of its block, and _deleting_ the temporary flag. 3. Thread 1 is (finally) scheduled again, and marks `s` as temporary with its transaction. 4. a) The process is killed, or thread 1's block fails verification and the temp flag is not deleted. This is a permanent failure! Any attempt to load state `s` will fail... hope it isn't on the main chain! Alternatively (4b) happens... b) Thread 1 finishes, and re-deletes the temporary flag. In this case the failure is transient, state `s` will disappear temporarily, but will come back once thread 1 finishes running. I _hope_ that steps 1-3 only happen very rarely, and 4a even more rarely. It's hard to know This once again begs the question of why we're using LevelDB (#483), when it clearly doesn't care about atomicity! A ham-fisted fix would be to wrap the hot and cold DBs in locks, which would bring us closer to how other DBs handle read-write transactions. E.g. [LMDB only allows one R/W transaction at a time](https://docs.rs/lmdb/0.8.0/lmdb/struct.Environment.html#method.begin_rw_txn). ### Race 2: Temporary state returned from `get_state` I don't think this race really matters, but in `load_hot_state`, if another thread stores a state between when we call `load_state_temporary_flag` and when we call `load_hot_state_summary`, then we could end up returning that state even though it's only a temporary state. I can't think of any case where this would be relevant, and I suspect if it did come up, it would be safe/recoverable (having data is safer than _not_ having data). This could be fixed by using a LevelDB read snapshot, but that would require substantial changes to how we read all our values, so I don't think it's worth it right now.
2020-10-23 01:27:51 +00:00
// On startup, the store should garbage collect all the temporary states.
let store = get_store(&db_path);
assert_eq!(store.iter_temporary_state_roots().count(), 0);
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn weak_subjectivity_sync() {
// Build an initial chain on one harness, representing a synced node with full history.
let num_initial_blocks = E::slots_per_epoch() * 11;
let num_final_blocks = E::slots_per_epoch() * 2;
let temp1 = tempdir().unwrap();
let full_store = get_store(&temp1);
let harness = get_harness(full_store.clone(), LOW_VALIDATOR_COUNT);
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
num_initial_blocks as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
let genesis_state = full_store
.get_state(&harness.chain.genesis_state_root, Some(Slot::new(0)))
.unwrap()
.unwrap();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let wss_checkpoint = harness.finalized_checkpoint();
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
let wss_block = harness
.chain
.store
.get_full_block(&wss_checkpoint.root)
.unwrap()
.unwrap();
let wss_state = full_store
.get_state(&wss_block.state_root(), None)
.unwrap()
.unwrap();
let wss_slot = wss_block.slot();
// Add more blocks that advance finalization further.
harness.advance_slot();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
num_final_blocks as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
let (shutdown_tx, _shutdown_rx) = futures::channel::mpsc::channel(1);
let log = test_logger();
let temp2 = tempdir().unwrap();
let store = get_store(&temp2);
let spec = test_spec::<E>();
let seconds_per_slot = spec.seconds_per_slot;
// Initialise a new beacon chain from the finalized checkpoint
let beacon_chain = Arc::new(
BeaconChainBuilder::new(MinimalEthSpec)
.store(store.clone())
.custom_spec(test_spec::<E>())
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
.task_executor(harness.chain.task_executor.clone())
.weak_subjectivity_state(wss_state, wss_block.clone(), genesis_state)
.unwrap()
.logger(log.clone())
.store_migrator_config(MigratorConfig::default().blocking())
.dummy_eth1_backend()
.expect("should build dummy backend")
.testing_slot_clock(Duration::from_secs(seconds_per_slot))
.expect("should configure testing slot clock")
.shutdown_sender(shutdown_tx)
.chain_config(ChainConfig::default())
.event_handler(Some(ServerSentEventHandler::new_with_capacity(
log.clone(),
1,
)))
.monitor_validators(true, vec![], log)
.build()
.expect("should build"),
);
// Apply blocks forward to reach head.
let chain_dump = harness.chain.chain_dump().unwrap();
let new_blocks = &chain_dump[wss_slot.as_usize() + 1..];
assert_eq!(new_blocks[0].beacon_block.slot(), wss_slot + 1);
for snapshot in new_blocks {
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
let full_block = harness
.chain
Prune finalized execution payloads (#3565) ## Issue Addressed Closes https://github.com/sigp/lighthouse/issues/3556 ## Proposed Changes Delete finalized execution payloads from the database in two places: 1. When running the finalization migration in `migrate_database`. We delete the finalized payloads between the last split point and the new updated split point. _If_ payloads are already pruned prior to this then this is sufficient to prune _all_ payloads as non-canonical payloads are already deleted by the head pruner, and all canonical payloads prior to the previous split will already have been pruned. 2. To address the fact that users will update to this code _after_ the merge on mainnet (and testnets), we need a one-off scan to delete the finalized payloads from the canonical chain. This is implemented in `try_prune_execution_payloads` which runs on startup and scans the chain back to the Bellatrix fork or the anchor slot (if checkpoint synced after Bellatrix). In the case where payloads are already pruned this check only imposes a single state load for the split state, which shouldn't be _too slow_. Even so, a flag `--prepare-payloads-on-startup=false` is provided to turn this off after it has run the first time, which provides faster start-up times. There is also a new `lighthouse db prune_payloads` subcommand for users who prefer to run the pruning manually. ## Additional Info The tests have been updated to not rely on finalized payloads in the database, instead using the `MockExecutionLayer` to reconstruct them. Additionally a check was added to `check_chain_dump` which asserts the non-existence or existence of payloads on disk depending on their slot.
2022-09-17 02:27:01 +00:00
.get_block(&snapshot.beacon_block_root)
.await
.unwrap()
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
.unwrap();
Prune finalized execution payloads (#3565) ## Issue Addressed Closes https://github.com/sigp/lighthouse/issues/3556 ## Proposed Changes Delete finalized execution payloads from the database in two places: 1. When running the finalization migration in `migrate_database`. We delete the finalized payloads between the last split point and the new updated split point. _If_ payloads are already pruned prior to this then this is sufficient to prune _all_ payloads as non-canonical payloads are already deleted by the head pruner, and all canonical payloads prior to the previous split will already have been pruned. 2. To address the fact that users will update to this code _after_ the merge on mainnet (and testnets), we need a one-off scan to delete the finalized payloads from the canonical chain. This is implemented in `try_prune_execution_payloads` which runs on startup and scans the chain back to the Bellatrix fork or the anchor slot (if checkpoint synced after Bellatrix). In the case where payloads are already pruned this check only imposes a single state load for the split state, which shouldn't be _too slow_. Even so, a flag `--prepare-payloads-on-startup=false` is provided to turn this off after it has run the first time, which provides faster start-up times. There is also a new `lighthouse db prune_payloads` subcommand for users who prefer to run the pruning manually. ## Additional Info The tests have been updated to not rely on finalized payloads in the database, instead using the `MockExecutionLayer` to reconstruct them. Additionally a check was added to `check_chain_dump` which asserts the non-existence or existence of payloads on disk depending on their slot.
2022-09-17 02:27:01 +00:00
let slot = full_block.slot();
let state_root = full_block.state_root();
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
Prune finalized execution payloads (#3565) ## Issue Addressed Closes https://github.com/sigp/lighthouse/issues/3556 ## Proposed Changes Delete finalized execution payloads from the database in two places: 1. When running the finalization migration in `migrate_database`. We delete the finalized payloads between the last split point and the new updated split point. _If_ payloads are already pruned prior to this then this is sufficient to prune _all_ payloads as non-canonical payloads are already deleted by the head pruner, and all canonical payloads prior to the previous split will already have been pruned. 2. To address the fact that users will update to this code _after_ the merge on mainnet (and testnets), we need a one-off scan to delete the finalized payloads from the canonical chain. This is implemented in `try_prune_execution_payloads` which runs on startup and scans the chain back to the Bellatrix fork or the anchor slot (if checkpoint synced after Bellatrix). In the case where payloads are already pruned this check only imposes a single state load for the split state, which shouldn't be _too slow_. Even so, a flag `--prepare-payloads-on-startup=false` is provided to turn this off after it has run the first time, which provides faster start-up times. There is also a new `lighthouse db prune_payloads` subcommand for users who prefer to run the pruning manually. ## Additional Info The tests have been updated to not rely on finalized payloads in the database, instead using the `MockExecutionLayer` to reconstruct them. Additionally a check was added to `check_chain_dump` which asserts the non-existence or existence of payloads on disk depending on their slot.
2022-09-17 02:27:01 +00:00
beacon_chain.slot_clock.set_slot(slot.as_u64());
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
beacon_chain
.process_block(
full_block.canonical_root(),
Arc::new(full_block),
CountUnrealized::True,
NotifyExecutionLayer::Yes,
)
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
.await
.unwrap();
beacon_chain.recompute_head_at_current_slot().await;
// Check that the new block's state can be loaded correctly.
let mut state = beacon_chain
.store
Prune finalized execution payloads (#3565) ## Issue Addressed Closes https://github.com/sigp/lighthouse/issues/3556 ## Proposed Changes Delete finalized execution payloads from the database in two places: 1. When running the finalization migration in `migrate_database`. We delete the finalized payloads between the last split point and the new updated split point. _If_ payloads are already pruned prior to this then this is sufficient to prune _all_ payloads as non-canonical payloads are already deleted by the head pruner, and all canonical payloads prior to the previous split will already have been pruned. 2. To address the fact that users will update to this code _after_ the merge on mainnet (and testnets), we need a one-off scan to delete the finalized payloads from the canonical chain. This is implemented in `try_prune_execution_payloads` which runs on startup and scans the chain back to the Bellatrix fork or the anchor slot (if checkpoint synced after Bellatrix). In the case where payloads are already pruned this check only imposes a single state load for the split state, which shouldn't be _too slow_. Even so, a flag `--prepare-payloads-on-startup=false` is provided to turn this off after it has run the first time, which provides faster start-up times. There is also a new `lighthouse db prune_payloads` subcommand for users who prefer to run the pruning manually. ## Additional Info The tests have been updated to not rely on finalized payloads in the database, instead using the `MockExecutionLayer` to reconstruct them. Additionally a check was added to `check_chain_dump` which asserts the non-existence or existence of payloads on disk depending on their slot.
2022-09-17 02:27:01 +00:00
.get_state(&state_root, Some(slot))
.unwrap()
.unwrap();
assert_eq!(state.update_tree_hash_cache().unwrap(), state_root);
}
// Forwards iterator from 0 should fail as we lack blocks.
assert!(matches!(
beacon_chain.forwards_iter_block_roots(Slot::new(0)),
Err(BeaconChainError::HistoricalBlockError(
HistoricalBlockError::BlockOutOfRange { .. }
))
));
// Simulate processing of a `StatusMessage` with an older finalized epoch by calling
// `block_root_at_slot` with an old slot for which we don't know the block root. It should
// return `None` rather than erroring.
assert_eq!(
beacon_chain
.block_root_at_slot(Slot::new(1), WhenSlotSkipped::None)
.unwrap(),
None
);
// Simulate querying the API for a historic state that is unknown. It should also return
// `None` rather than erroring.
assert_eq!(beacon_chain.state_root_at_slot(Slot::new(1)).unwrap(), None);
// Supply blocks backwards to reach genesis. Omit the genesis block to check genesis handling.
let historical_blocks = chain_dump[..wss_block.slot().as_usize()]
.iter()
.filter(|s| s.beacon_block.slot() != 0)
.map(|s| s.beacon_block.clone())
.collect::<Vec<_>>();
beacon_chain
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
.import_historical_block_batch(historical_blocks.clone())
.unwrap();
assert_eq!(beacon_chain.store.get_oldest_block_slot(), 0);
// Resupplying the blocks should not fail, they can be safely ignored.
beacon_chain
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
.import_historical_block_batch(historical_blocks)
.unwrap();
// The forwards iterator should now match the original chain
let forwards = beacon_chain
.forwards_iter_block_roots(Slot::new(0))
.unwrap()
.map(Result::unwrap)
.collect::<Vec<_>>();
let expected = harness
.chain
.forwards_iter_block_roots(Slot::new(0))
.unwrap()
.map(Result::unwrap)
.collect::<Vec<_>>();
assert_eq!(forwards, expected);
// All blocks can be loaded.
for (block_root, slot) in beacon_chain
.forwards_iter_block_roots(Slot::new(0))
.unwrap()
.map(Result::unwrap)
{
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
let block = store.get_blinded_block(&block_root).unwrap().unwrap();
assert_eq!(block.slot(), slot);
}
// All states from the oldest state slot can be loaded.
let (_, oldest_state_slot) = store.get_historic_state_limits();
for (state_root, slot) in beacon_chain
.forwards_iter_state_roots(oldest_state_slot)
.unwrap()
.map(Result::unwrap)
{
let state = store.get_state(&state_root, Some(slot)).unwrap().unwrap();
assert_eq!(state.slot(), slot);
assert_eq!(state.canonical_root(), state_root);
}
// Anchor slot is still set to the starting slot.
assert_eq!(store.get_anchor_slot(), Some(wss_slot));
// Reconstruct states.
store.clone().reconstruct_historic_states().unwrap();
assert_eq!(store.get_anchor_slot(), None);
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn finalizes_after_resuming_from_db() {
let validator_count = 16;
let num_blocks_produced = MinimalEthSpec::slots_per_epoch() * 8;
let first_half = num_blocks_produced / 2;
let db_path = tempdir().unwrap();
let store = get_store(&db_path);
let harness = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.keypairs(KEYPAIRS[0..validator_count].to_vec())
.fresh_disk_store(store.clone())
.mock_execution_layer()
.build();
harness.advance_slot();
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
harness
.extend_chain(
first_half as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
assert!(
harness
.chain
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
.head_snapshot()
.beacon_state
.finalized_checkpoint()
.epoch
> 0,
"the chain should have already finalized"
);
let latest_slot = harness.chain.slot().expect("should have a slot");
harness
.chain
.persist_head_and_fork_choice()
.expect("should persist the head and fork choice");
harness
.chain
.persist_op_pool()
.expect("should persist the op pool");
harness
.chain
.persist_eth1_cache()
.expect("should persist the eth1 cache");
let original_chain = harness.chain;
let resumed_harness = BeaconChainHarness::builder(MinimalEthSpec)
.default_spec()
.keypairs(KEYPAIRS[0..validator_count].to_vec())
.resumed_disk_store(store)
.mock_execution_layer()
.build();
assert_chains_pretty_much_the_same(&original_chain, &resumed_harness.chain);
// Set the slot clock of the resumed harness to be in the slot following the previous harness.
//
// This allows us to produce the block at the next slot.
resumed_harness
.chain
.slot_clock
.set_slot(latest_slot.as_u64() + 1);
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
resumed_harness
.extend_chain(
(num_blocks_produced - first_half) as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
.await;
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let state = &resumed_harness.chain.head_snapshot().beacon_state;
assert_eq!(
state.slot(),
num_blocks_produced,
"head should be at the current slot"
);
assert_eq!(
state.current_epoch(),
num_blocks_produced / MinimalEthSpec::slots_per_epoch(),
"head should be at the expected epoch"
);
assert_eq!(
state.current_justified_checkpoint().epoch,
state.current_epoch() - 1,
"the head should be justified one behind the current epoch"
);
assert_eq!(
state.finalized_checkpoint().epoch,
state.current_epoch() - 2,
"the head should be finalized two behind the current epoch"
);
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
#[tokio::test]
async fn revert_minority_fork_on_resume() {
let validator_count = 16;
let slots_per_epoch = MinimalEthSpec::slots_per_epoch();
let fork_epoch = Epoch::new(4);
let fork_slot = fork_epoch.start_slot(slots_per_epoch);
let initial_blocks = slots_per_epoch * fork_epoch.as_u64() - 1;
let post_fork_blocks = slots_per_epoch * 3;
let mut spec1 = MinimalEthSpec::default_spec();
spec1.altair_fork_epoch = None;
let mut spec2 = MinimalEthSpec::default_spec();
spec2.altair_fork_epoch = Some(fork_epoch);
let seconds_per_slot = spec1.seconds_per_slot;
let all_validators = (0..validator_count).collect::<Vec<usize>>();
// Chain with no fork epoch configured.
let db_path1 = tempdir().unwrap();
let store1 = get_store_with_spec(&db_path1, spec1.clone());
let harness1 = BeaconChainHarness::builder(MinimalEthSpec)
.spec(spec1)
.keypairs(KEYPAIRS[0..validator_count].to_vec())
.fresh_disk_store(store1)
.mock_execution_layer()
.build();
// Chain with fork epoch configured.
let db_path2 = tempdir().unwrap();
let store2 = get_store_with_spec(&db_path2, spec2.clone());
let harness2 = BeaconChainHarness::builder(MinimalEthSpec)
.spec(spec2.clone())
.keypairs(KEYPAIRS[0..validator_count].to_vec())
.fresh_disk_store(store2)
.mock_execution_layer()
.build();
// Apply the same blocks to both chains initially.
let mut state = harness1.get_current_state();
let mut block_root = harness1.chain.genesis_block_root;
for slot in (1..=initial_blocks).map(Slot::new) {
let state_root = state.update_tree_hash_cache().unwrap();
let attestations = harness1.make_attestations(
&all_validators,
&state,
state_root,
block_root.into(),
slot,
);
harness1.set_current_slot(slot);
harness2.set_current_slot(slot);
harness1.process_attestations(attestations.clone());
harness2.process_attestations(attestations);
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (block, new_state) = harness1.make_block(state, slot).await;
harness1
.process_block(slot, block.canonical_root(), block.clone())
.await
.unwrap();
harness2
.process_block(slot, block.canonical_root(), block.clone())
.await
.unwrap();
state = new_state;
block_root = block.canonical_root();
}
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
assert_eq!(harness1.head_slot(), fork_slot - 1);
assert_eq!(harness2.head_slot(), fork_slot - 1);
// Fork the two chains.
let mut state1 = state.clone();
let mut state2 = state.clone();
let mut majority_blocks = vec![];
for i in 0..post_fork_blocks {
let slot = fork_slot + i;
// Attestations on majority chain.
let state_root = state.update_tree_hash_cache().unwrap();
let attestations = harness2.make_attestations(
&all_validators,
&state2,
state_root,
block_root.into(),
slot,
);
harness2.set_current_slot(slot);
harness2.process_attestations(attestations);
// Minority chain block (no attesters).
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (block1, new_state1) = harness1.make_block(state1, slot).await;
harness1
.process_block(slot, block1.canonical_root(), block1)
.await
.unwrap();
state1 = new_state1;
// Majority chain block (all attesters).
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let (block2, new_state2) = harness2.make_block(state2, slot).await;
harness2
.process_block(slot, block2.canonical_root(), block2.clone())
.await
.unwrap();
state2 = new_state2;
block_root = block2.canonical_root();
majority_blocks.push(block2);
}
let end_slot = fork_slot + post_fork_blocks - 1;
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
assert_eq!(harness1.head_slot(), end_slot);
assert_eq!(harness2.head_slot(), end_slot);
// Resume from disk with the hard-fork activated: this should revert the post-fork blocks.
// We have to do some hackery with the `slot_clock` so that the correct slot is set when
// the beacon chain builder loads the head block.
drop(harness1);
let resume_store = get_store_with_spec(&db_path1, spec2.clone());
let resumed_harness = BeaconChainHarness::builder(MinimalEthSpec)
.spec(spec2)
.keypairs(KEYPAIRS[0..validator_count].to_vec())
.resumed_disk_store(resume_store)
.override_store_mutator(Box::new(move |mut builder| {
builder = builder
.resume_from_db()
.unwrap()
.testing_slot_clock(Duration::from_secs(seconds_per_slot))
.unwrap();
builder
.get_slot_clock()
.unwrap()
.set_slot(end_slot.as_u64());
builder
}))
.mock_execution_layer()
.build();
// Head should now be just before the fork.
resumed_harness.chain.recompute_head_at_current_slot().await;
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
assert_eq!(resumed_harness.head_slot(), fork_slot - 1);
// Head track should know the canonical head and the rogue head.
assert_eq!(resumed_harness.chain.heads().len(), 2);
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
assert!(resumed_harness
.chain
.knows_head(&resumed_harness.head_block_root().into()));
// Apply blocks from the majority chain and trigger finalization.
let initial_split_slot = resumed_harness.chain.store.get_split_slot();
for block in &majority_blocks {
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
resumed_harness
.process_block_result(block.clone())
.await
.unwrap();
// The canonical head should be the block from the majority chain.
resumed_harness.chain.recompute_head_at_current_slot().await;
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
assert_eq!(resumed_harness.head_slot(), block.slot());
assert_eq!(resumed_harness.head_block_root(), block.canonical_root());
}
let advanced_split_slot = resumed_harness.chain.store.get_split_slot();
// Check that the migration ran successfully.
assert!(advanced_split_slot > initial_split_slot);
// Check that there is only a single head now matching harness2 (the minority chain is gone).
let heads = resumed_harness.chain.heads();
assert_eq!(heads, harness2.chain.heads());
assert_eq!(heads.len(), 1);
}
/// Checks that two chains are the same, for the purpose of these tests.
///
/// Several fields that are hard/impossible to check are ignored (e.g., the store).
fn assert_chains_pretty_much_the_same<T: BeaconChainTypes>(a: &BeaconChain<T>, b: &BeaconChain<T>) {
assert_eq!(a.spec, b.spec, "spec should be equal");
assert_eq!(a.op_pool, b.op_pool, "op_pool should be equal");
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let a_head = a.head_snapshot();
let b_head = b.head_snapshot();
assert_eq!(
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
a_head.beacon_block_root, b_head.beacon_block_root,
"head block roots should be equal"
);
assert_eq!(
a_head.beacon_block, b_head.beacon_block,
"head blocks should be equal"
);
// Clone with committee caches only to prevent other caches from messing with the equality
// check.
assert_eq!(
a_head.beacon_state.clone_with_only_committee_caches(),
b_head.beacon_state.clone_with_only_committee_caches(),
"head states should be equal"
);
assert_eq!(a.heads(), b.heads(), "heads() should be equal");
assert_eq!(
a.genesis_block_root, b.genesis_block_root,
"genesis_block_root should be equal"
);
let slot = a.slot().unwrap();
v1.1.6 Fork Choice changes (#2822) ## Issue Addressed Resolves: https://github.com/sigp/lighthouse/issues/2741 Includes: https://github.com/sigp/lighthouse/pull/2853 so that we can get ssz static tests passing here on v1.1.6. If we want to merge that first, we can make this diff slightly smaller ## Proposed Changes - Changes the `justified_epoch` and `finalized_epoch` in the `ProtoArrayNode` each to an `Option<Checkpoint>`. The `Option` is necessary only for the migration, so not ideal. But does allow us to add a default logic to `None` on these fields during the database migration. - Adds a database migration from a legacy fork choice struct to the new one, search for all necessary block roots in fork choice by iterating through blocks in the db. - updates related to https://github.com/ethereum/consensus-specs/pull/2727 - We will have to update the persisted forkchoice to make sure the justified checkpoint stored is correct according to the updated fork choice logic. This boils down to setting the forkchoice store's justified checkpoint to the justified checkpoint of the block that advanced the finalized checkpoint to the current one. - AFAICT there's no migration steps necessary for the update to allow applying attestations from prior blocks, but would appreciate confirmation on that - I updated the consensus spec tests to v1.1.6 here, but they will fail until we also implement the proposer score boost updates. I confirmed that the previously failing scenario `new_finalized_slot_is_justified_checkpoint_ancestor` will now pass after the boost updates, but haven't confirmed _all_ tests will pass because I just quickly stubbed out the proposer boost test scenario formatting. - This PR now also includes proposer boosting https://github.com/ethereum/consensus-specs/pull/2730 ## Additional Info I realized checking justified and finalized roots in fork choice makes it more likely that we trigger this bug: https://github.com/ethereum/consensus-specs/pull/2727 It's possible the combination of justified checkpoint and finalized checkpoint in the forkchoice store is different from in any block in fork choice. So when trying to startup our store's justified checkpoint seems invalid to the rest of fork choice (but it should be valid). When this happens we get an `InvalidBestNode` error and fail to start up. So I'm including that bugfix in this branch. Todo: - [x] Fix fork choice tests - [x] Self review - [x] Add fix for https://github.com/ethereum/consensus-specs/pull/2727 - [x] Rebase onto Kintusgi - [x] Fix `num_active_validators` calculation as @michaelsproul pointed out - [x] Clean up db migrations Co-authored-by: realbigsean <seananderson33@gmail.com>
2021-12-13 20:43:22 +00:00
let spec = T::EthSpec::default_spec();
assert!(
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
a.canonical_head
.fork_choice_write_lock()
.get_head(slot, &spec)
.unwrap()
== b.canonical_head
.fork_choice_write_lock()
.get_head(slot, &spec)
.unwrap(),
"fork_choice heads should be equal"
);
}
/// Check that the head state's slot matches `expected_slot`.
fn check_slot(harness: &TestHarness, expected_slot: u64) {
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let state = &harness.chain.head_snapshot().beacon_state;
assert_eq!(
state.slot(),
expected_slot,
"head should be at the current slot"
);
}
/// Check that the chain has finalized under best-case assumptions, and check the head slot.
fn check_finalization(harness: &TestHarness, expected_slot: u64) {
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
let state = &harness.chain.head_snapshot().beacon_state;
check_slot(harness, expected_slot);
assert_eq!(
state.current_justified_checkpoint().epoch,
state.current_epoch() - 1,
"the head should be justified one behind the current epoch"
);
assert_eq!(
state.finalized_checkpoint().epoch,
state.current_epoch() - 2,
"the head should be finalized two behind the current epoch"
);
}
/// Check that the HotColdDB's split_slot is equal to the start slot of the last finalized epoch.
fn check_split_slot(harness: &TestHarness, store: Arc<HotColdDB<E, LevelDB<E>, LevelDB<E>>>) {
let split_slot = store.get_split_slot();
assert_eq!(
harness
.chain
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
.head_snapshot()
.beacon_state
.finalized_checkpoint()
.epoch
.start_slot(E::slots_per_epoch()),
split_slot
);
assert_ne!(split_slot, 0);
}
/// Check that all the states in a chain dump have the correct tree hash.
fn check_chain_dump(harness: &TestHarness, expected_len: u64) {
let chain_dump = harness.chain.chain_dump().unwrap();
Prune finalized execution payloads (#3565) ## Issue Addressed Closes https://github.com/sigp/lighthouse/issues/3556 ## Proposed Changes Delete finalized execution payloads from the database in two places: 1. When running the finalization migration in `migrate_database`. We delete the finalized payloads between the last split point and the new updated split point. _If_ payloads are already pruned prior to this then this is sufficient to prune _all_ payloads as non-canonical payloads are already deleted by the head pruner, and all canonical payloads prior to the previous split will already have been pruned. 2. To address the fact that users will update to this code _after_ the merge on mainnet (and testnets), we need a one-off scan to delete the finalized payloads from the canonical chain. This is implemented in `try_prune_execution_payloads` which runs on startup and scans the chain back to the Bellatrix fork or the anchor slot (if checkpoint synced after Bellatrix). In the case where payloads are already pruned this check only imposes a single state load for the split state, which shouldn't be _too slow_. Even so, a flag `--prepare-payloads-on-startup=false` is provided to turn this off after it has run the first time, which provides faster start-up times. There is also a new `lighthouse db prune_payloads` subcommand for users who prefer to run the pruning manually. ## Additional Info The tests have been updated to not rely on finalized payloads in the database, instead using the `MockExecutionLayer` to reconstruct them. Additionally a check was added to `check_chain_dump` which asserts the non-existence or existence of payloads on disk depending on their slot.
2022-09-17 02:27:01 +00:00
let split_slot = harness.chain.store.get_split_slot();
assert_eq!(chain_dump.len() as u64, expected_len);
for checkpoint in &chain_dump {
// Check that the tree hash of the stored state is as expected
assert_eq!(
Advance state to next slot after importing block (#2174) ## Issue Addressed NA ## Proposed Changes Add an optimization to perform `per_slot_processing` from the *leading-edge* of block processing to the *trailing-edge*. Ultimately, this allows us to import the block at slot `n` faster because we used the tail-end of slot `n - 1` to perform `per_slot_processing`. Additionally, add a "block proposer cache" which allows us to cache the block proposer for some epoch. Since we're now doing trailing-edge `per_slot_processing`, we can prime this cache with the values for the next epoch before those blocks arrive (assuming those blocks don't have some weird forking). There were several ancillary changes required to achieve this: - Remove the `state_root` field of `BeaconSnapshot`, since there's no need to know it on a `pre_state` and in all other cases we can just read it from `block.state_root()`. - This caused some "dust" changes of `snapshot.beacon_state_root` to `snapshot.beacon_state_root()`, where the `BeaconSnapshot::beacon_state_root()` func just reads the state root from the block. - Rename `types::ShuffingId` to `AttestationShufflingId`. I originally did this because I added a `ProposerShufflingId` struct which turned out to be not so useful. I thought this new name was more descriptive so I kept it. - Address https://github.com/ethereum/eth2.0-specs/pull/2196 - Add a debug log when we get a block with an unknown parent. There was previously no logging around this case. - Add a function to `BeaconState` to compute all proposers for an epoch without re-computing the active indices for each slot. ## Additional Info - ~~Blocked on #2173~~ - ~~Blocked on #2179~~ That PR was wrapped into this PR. - There's potentially some places where we could avoid computing the proposer indices in `per_block_processing` but I haven't done this here. These would be an optimization beyond the issue at hand (improving block propagation times) and I think this PR is already doing enough. We can come back for that later. ## TODO - [x] Tidy, improve comments. - [x] ~~Try avoid computing proposer index in `per_block_processing`?~~
2021-02-15 07:17:52 +00:00
checkpoint.beacon_state_root(),
checkpoint.beacon_state.tree_hash_root(),
"tree hash of stored state is incorrect"
);
// Check that looking up the state root with no slot hint succeeds.
// This tests the state root -> slot mapping.
assert_eq!(
harness
.chain
.store
Advance state to next slot after importing block (#2174) ## Issue Addressed NA ## Proposed Changes Add an optimization to perform `per_slot_processing` from the *leading-edge* of block processing to the *trailing-edge*. Ultimately, this allows us to import the block at slot `n` faster because we used the tail-end of slot `n - 1` to perform `per_slot_processing`. Additionally, add a "block proposer cache" which allows us to cache the block proposer for some epoch. Since we're now doing trailing-edge `per_slot_processing`, we can prime this cache with the values for the next epoch before those blocks arrive (assuming those blocks don't have some weird forking). There were several ancillary changes required to achieve this: - Remove the `state_root` field of `BeaconSnapshot`, since there's no need to know it on a `pre_state` and in all other cases we can just read it from `block.state_root()`. - This caused some "dust" changes of `snapshot.beacon_state_root` to `snapshot.beacon_state_root()`, where the `BeaconSnapshot::beacon_state_root()` func just reads the state root from the block. - Rename `types::ShuffingId` to `AttestationShufflingId`. I originally did this because I added a `ProposerShufflingId` struct which turned out to be not so useful. I thought this new name was more descriptive so I kept it. - Address https://github.com/ethereum/eth2.0-specs/pull/2196 - Add a debug log when we get a block with an unknown parent. There was previously no logging around this case. - Add a function to `BeaconState` to compute all proposers for an epoch without re-computing the active indices for each slot. ## Additional Info - ~~Blocked on #2173~~ - ~~Blocked on #2179~~ That PR was wrapped into this PR. - There's potentially some places where we could avoid computing the proposer indices in `per_block_processing` but I haven't done this here. These would be an optimization beyond the issue at hand (improving block propagation times) and I think this PR is already doing enough. We can come back for that later. ## TODO - [x] Tidy, improve comments. - [x] ~~Try avoid computing proposer index in `per_block_processing`?~~
2021-02-15 07:17:52 +00:00
.get_state(&checkpoint.beacon_state_root(), None)
.expect("no error")
.expect("state exists")
.slot(),
checkpoint.beacon_state.slot()
);
Prune finalized execution payloads (#3565) ## Issue Addressed Closes https://github.com/sigp/lighthouse/issues/3556 ## Proposed Changes Delete finalized execution payloads from the database in two places: 1. When running the finalization migration in `migrate_database`. We delete the finalized payloads between the last split point and the new updated split point. _If_ payloads are already pruned prior to this then this is sufficient to prune _all_ payloads as non-canonical payloads are already deleted by the head pruner, and all canonical payloads prior to the previous split will already have been pruned. 2. To address the fact that users will update to this code _after_ the merge on mainnet (and testnets), we need a one-off scan to delete the finalized payloads from the canonical chain. This is implemented in `try_prune_execution_payloads` which runs on startup and scans the chain back to the Bellatrix fork or the anchor slot (if checkpoint synced after Bellatrix). In the case where payloads are already pruned this check only imposes a single state load for the split state, which shouldn't be _too slow_. Even so, a flag `--prepare-payloads-on-startup=false` is provided to turn this off after it has run the first time, which provides faster start-up times. There is also a new `lighthouse db prune_payloads` subcommand for users who prefer to run the pruning manually. ## Additional Info The tests have been updated to not rely on finalized payloads in the database, instead using the `MockExecutionLayer` to reconstruct them. Additionally a check was added to `check_chain_dump` which asserts the non-existence or existence of payloads on disk depending on their slot.
2022-09-17 02:27:01 +00:00
// Check presence of execution payload on disk.
if harness.chain.spec.bellatrix_fork_epoch.is_some() {
assert_eq!(
harness
.chain
.store
.execution_payload_exists(&checkpoint.beacon_block_root)
.unwrap(),
checkpoint.beacon_block.slot() >= split_slot,
"incorrect payload storage for block at slot {}: {:?}",
checkpoint.beacon_block.slot(),
checkpoint.beacon_block_root,
);
}
}
// Check the forwards block roots iterator against the chain dump
let chain_dump_block_roots = chain_dump
.iter()
Update to Spec v0.10 (#817) * Start updating types * WIP * Signature hacking * Existing EF tests passing with fake_crypto * Updates * Delete outdated API spec * The refactor continues * It compiles * WIP test fixes * All release tests passing bar genesis state parsing * Update and test YamlConfig * Update to spec v0.10 compatible BLS * Updates to BLS EF tests * Add EF test for AggregateVerify And delete unused hash2curve tests for uncompressed points * Update EF tests to v0.10.1 * Use optional block root correctly in block proc * Use genesis fork in deposit domain. All tests pass * Cargo fmt * Fast aggregate verify test * Update REST API docs * Cargo fmt * Fix unused import * Bump spec tags to v0.10.1 * Add `seconds_per_eth1_block` to chainspec * Update to timestamp based eth1 voting scheme * Return None from `get_votes_to_consider` if block cache is empty * Handle overflows in `is_candidate_block` * Revert to failing tests * Fix eth1 data sets test * Choose default vote according to spec * Fix collect_valid_votes tests * Fix `get_votes_to_consider` to choose all eligible blocks * Uncomment winning_vote tests * Add comments; remove unused code * Reduce seconds_per_eth1_block for simulation * Addressed review comments * Add test for default vote case * Fix logs * Remove unused functions * Meter default eth1 votes * Fix comments * Address review comments; remove unused dependency * Disable/delete two outdated tests * Bump eth1 default vote warn to error * Delete outdated eth1 test Co-authored-by: Pawan Dhananjay <pawandhananjay@gmail.com>
2020-02-10 23:19:36 +00:00
.map(|checkpoint| (checkpoint.beacon_block_root, checkpoint.beacon_block.slot()))
.collect::<Vec<_>>();
let mut forward_block_roots = harness
.chain
.forwards_iter_block_roots(Slot::new(0))
.expect("should get iter")
.map(Result::unwrap)
.collect::<Vec<_>>();
// Drop the block roots for skipped slots.
forward_block_roots.dedup_by_key(|(block_root, _)| *block_root);
for i in 0..std::cmp::max(chain_dump_block_roots.len(), forward_block_roots.len()) {
assert_eq!(
chain_dump_block_roots[i],
forward_block_roots[i],
"split slot is {}",
harness.chain.store.get_split_slot()
);
}
}
/// Check that every state from the canonical chain is in the database, and that the
/// reverse state and block root iterators reach genesis.
fn check_iterators(harness: &TestHarness) {
let mut max_slot = None;
for (state_root, slot) in harness
.chain
.forwards_iter_state_roots(Slot::new(0))
.expect("should get iter")
.map(Result::unwrap)
{
assert!(
harness
.chain
.store
.get_state(&state_root, Some(slot))
.unwrap()
.is_some(),
"state {:?} from canonical chain should be in DB",
state_root
);
max_slot = Some(slot);
}
// Assert that we reached the head.
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
assert_eq!(max_slot, Some(harness.head_slot()));
// Assert that the block root iterator reaches the head.
assert_eq!(
harness
.chain
.forwards_iter_block_roots(Slot::new(0))
.expect("should get iter")
.last()
.map(Result::unwrap)
.map(|(_, slot)| slot),
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
Some(harness.head_slot())
);
}
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
fn get_finalized_epoch_boundary_blocks(
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
dump: &[BeaconSnapshot<MinimalEthSpec, BlindedPayload<MinimalEthSpec>>],
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
) -> HashSet<SignedBeaconBlockHash> {
dump.iter()
.cloned()
.map(|checkpoint| checkpoint.beacon_state.finalized_checkpoint().root.into())
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
.collect()
}
Separate execution payloads in the DB (#3157) ## Proposed Changes Reduce post-merge disk usage by not storing finalized execution payloads in Lighthouse's database. :warning: **This is achieved in a backwards-incompatible way for networks that have already merged** :warning:. Kiln users and shadow fork enjoyers will be unable to downgrade after running the code from this PR. The upgrade migration may take several minutes to run, and can't be aborted after it begins. The main changes are: - New column in the database called `ExecPayload`, keyed by beacon block root. - The `BeaconBlock` column now stores blinded blocks only. - Lots of places that previously used full blocks now use blinded blocks, e.g. analytics APIs, block replay in the DB, etc. - On finalization: - `prune_abanonded_forks` deletes non-canonical payloads whilst deleting non-canonical blocks. - `migrate_db` deletes finalized canonical payloads whilst deleting finalized states. - Conversions between blinded and full blocks are implemented in a compositional way, duplicating some work from Sean's PR #3134. - The execution layer has a new `get_payload_by_block_hash` method that reconstructs a payload using the EE's `eth_getBlockByHash` call. - I've tested manually that it works on Kiln, using Geth and Nethermind. - This isn't necessarily the most efficient method, and new engine APIs are being discussed to improve this: https://github.com/ethereum/execution-apis/pull/146. - We're depending on the `ethers` master branch, due to lots of recent changes. We're also using a workaround for https://github.com/gakonst/ethers-rs/issues/1134. - Payload reconstruction is used in the HTTP API via `BeaconChain::get_block`, which is now `async`. Due to the `async` fn, the `blocking_json` wrapper has been removed. - Payload reconstruction is used in network RPC to serve blocks-by-{root,range} responses. Here the `async` adjustment is messier, although I think I've managed to come up with a reasonable compromise: the handlers take the `SendOnDrop` by value so that they can drop it on _task completion_ (after the `fn` returns). Still, this is introducing disk reads onto core executor threads, which may have a negative performance impact (thoughts appreciated). ## Additional Info - [x] For performance it would be great to remove the cloning of full blocks when converting them to blinded blocks to write to disk. I'm going to experiment with a `put_block` API that takes the block by value, breaks it into a blinded block and a payload, stores the blinded block, and then re-assembles the full block for the caller. - [x] We should measure the latency of blocks-by-root and blocks-by-range responses. - [x] We should add integration tests that stress the payload reconstruction (basic tests done, issue for more extensive tests: https://github.com/sigp/lighthouse/issues/3159) - [x] We should (manually) test the schema v9 migration from several prior versions, particularly as blocks have changed on disk and some migrations rely on being able to load blocks. Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-05-12 00:42:17 +00:00
fn get_blocks(
dump: &[BeaconSnapshot<MinimalEthSpec, BlindedPayload<MinimalEthSpec>>],
) -> HashSet<SignedBeaconBlockHash> {
Prune abandoned forks (#916) * Address compiler warning * Prune abandoned fork choice forks * New approach to pruning * Wrap some block hashes in a newtype pattern For increased type safety. * Add Graphviz chain dump emitter for debugging * Fix broken test case * Make prunes_abandoned_forks use real DiskStore * Mark finalized blocks in the GraphViz output * Refine debug stringification of Slot and Epoch Before this commit: print!("{:?}", Slot(123)) == "Slot(\n123\n)". After this commit: print!("{:?", Slot(123)) == "Slot(123)". * Simplify build_block() * Rewrite test case using more composable test primitives * Working rewritten test case * Tighten fork prunning test checks * Add another pruning test case * Bugfix: Finalized blocks weren't always properly detected * Pruning: Add pruning_does_not_touch_blocks_prior_to_finalization test case * Tighten pruning tests: check if heads are tracked properly * Add a failing test case for a buggy scenario * Change name of function to a more accurate one * Fix failing test case * Test case: Were skipped slots' states pruned? * Style fix: Simplify dereferencing * Tighten pruning tests: check if abandoned states are deleted * Towards atomicity of db ops * Correct typo * Prune also skipped slots' states * New logic for handling skipped states * Make skipped slots test pass * Post conflict resolution fixes * Formatting fixes * Tests passing * Block hashes in Graphviz node labels * Removed unused changes * Fix bug with states having < SlotsPerHistoricalRoot roots * Consolidate State/BlockRootsIterator for pruning * Address review feedback * Fix a bug in pruning tests * Detach prune_abandoned_forks() from its object * Move migrate.rs from store to beacon_chain * Move forks pruning onto a background thread * Bugfix: Heads weren't pruned when prune set contained only the head * Rename: freeze_to_state() -> process_finalization() * Eliminate redundant function parameter Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-04-20 09:59:56 +00:00
dump.iter()
.cloned()
.map(|checkpoint| checkpoint.beacon_block_root.into())
.collect()
}