lighthouse/beacon_node/beacon_chain/src/test_utils.rs

1789 lines
63 KiB
Rust
Raw Normal View History

pub use crate::persisted_beacon_chain::PersistedBeaconChain;
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
pub use crate::{
beacon_chain::{BEACON_CHAIN_DB_KEY, ETH1_CACHE_DB_KEY, FORK_CHOICE_DB_KEY, OP_POOL_DB_KEY},
migrate::MigratorConfig,
BeaconChainError, ProduceBlockVerification,
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 crate::{
Eth1 Integration (#542) * Refactor to cache Eth1Data * Fix merge conflicts and minor refactorings * Rename Eth1Cache to Eth1DataCache * Refactor events subscription * Add deposits module to interface with BeaconChain deposits * Remove utils * Rename to types.rs and add trait constraints to Eth1DataFetcher * Confirm to trait constraints. Make Web3DataFetcher cloneable * Make fetcher object member of deposit and eth1_data cache and other fixes * Fix update_cache function * Move fetch_eth1_data to impl block * Fix deposit tests * Create Eth1 object for interfacing with Beacon chain * Add `run` function for running update_cache and subscribe_deposit_logs tasks * Add logging * Run `cargo fmt` and make tests pass * Convert sync functions to async * Add timeouts to web3 functions * Return futures from cache functions * Add failed chaining of futures * Working cache updation * Clean up tests and `update_cache` function * Refactor `get_eth1_data` functions to work with future returning functions * Refactor eth1 `run` function to work with modified `update_cache` api * Minor changes * Add distance parameter to `update_cache` * Fix tests and other minor fixes * Working integration with cache and deposits * Add merkle_tree construction, proof generation and verification code * Add function to construct and fetch Deposits for BeaconNode * Add error handling * Import ssz * Add error handling to eth1 cache and fix minor errors * Run rustfmt * Fix minor bug * Rename Eth1Error and change to Result<T> * Change deposit fetching mechanism from notification based to poll based * Add deposits from eth1 chain in a given range every `x` blocks * Modify `run` function to accommodate changes * Minor fixes * Fix formatting * Initial commit. web3 api working. * Tidied up lib. Add function for fetching logs. * Refactor with `Eth1DataFetcher` trait * Add parsing for deposit contract logs and get_eth1_data function * Add `get_eth1_votes` function * Refactor to cache Eth1Data * Fix merge conflicts and minor refactorings * Rename Eth1Cache to Eth1DataCache * Refactor events subscription * Add deposits module to interface with BeaconChain deposits * Remove utils * Rename to types.rs and add trait constraints to Eth1DataFetcher * Confirm to trait constraints. Make Web3DataFetcher cloneable * Make fetcher object member of deposit and eth1_data cache and other fixes * Fix update_cache function * Move fetch_eth1_data to impl block * Fix deposit tests * Create Eth1 object for interfacing with Beacon chain * Add `run` function for running update_cache and subscribe_deposit_logs tasks * Add logging * Run `cargo fmt` and make tests pass * Convert sync functions to async * Add timeouts to web3 functions * Return futures from cache functions * Add failed chaining of futures * Working cache updation * Clean up tests and `update_cache` function * Refactor `get_eth1_data` functions to work with future returning functions * Refactor eth1 `run` function to work with modified `update_cache` api * Minor changes * Add distance parameter to `update_cache` * Fix tests and other minor fixes * Working integration with cache and deposits * Add merkle_tree construction, proof generation and verification code * Add function to construct and fetch Deposits for BeaconNode * Add error handling * Import ssz * Add error handling to eth1 cache and fix minor errors * Run rustfmt * Fix minor bug * Rename Eth1Error and change to Result<T> * Change deposit fetching mechanism from notification based to poll based * Add deposits from eth1 chain in a given range every `x` blocks * Modify `run` function to accommodate changes * Minor fixes * Fix formatting * Fix merge issue * Refactor with `Config` struct. Remote `ContractConfig` * Rename eth1_chain crate to eth1 * Rename files and read abi file using `fs::read` * Move eth1 to lib * Remove unnecessary mutability constraint * Add `Web3Backend` for returning actual eth1 data * Refactor `get_eth1_votes` to return a Result * Delete `eth1_chain` crate * Return `Result` from `get_deposits` * Fix range of deposits to return to beacon chain * Add `get_block_height_by_hash` trait function * Add naive method for getting `previous_eth1_distance` * Add eth1 config params to main config * Add instructions for setting up eth1 testing environment * Add build script to fetch deposit contract abi * Contract ABI is part of compiled binary * Fix minor bugs * Move docs to lib * Add timeout to config * Remove print statements * Change warn to error * Fix typos * Removed prints in test and get timeout value from config * Fixed error types * Added logging to web3_fetcher * Refactor for modified web3 api * Fix minor stuff * Add build script * Tidy, hide eth1 integration tests behind flag * Add http crate * Add first stages of eth1_test_rig * Fix deposits on test rig * Fix bug with deposit count method * Add block hash getter to http eth1 * Clean eth1 http crate and tests * Add script to start ganache * Adds deposit tree to eth1-http * Extend deposit tree tests * Tidy tests in eth1-http * Add more detail to get block request * Add block cache to eth1-http * Rename deposit tree to deposit cache * Add inital updating to eth1-http * Tidy updater * Fix compile bugs in tests * Adds an Eth1DataCache builder * Reorg eth1-http files * Add (failing) tests for eth1 updater * Rename files, fix bug in eth1-http * Ensure that ganache timestamps are increasing * Fix bugs with getting eth1data ancestors * Improve eth1 testing, fix bugs * Add truncate method to block cache * Add pruning to block cache update process * Add tests for block pruning * Allow for dropping an expired cache. * Add more comments * Add first compiling version of deposit updater * Add common fn for getting range of required blocks * Add passing deposit update test * Improve tests * Fix block pruning bug * Add tests for running two updates at once * Add updater services to eth1 * Add deposit collection to beacon chain * Add incomplete builder experiments * Add first working version of beacon chain builder * Update test harness to new beacon chain type * Rename builder file, tidy * Add first working client builder * Progress further on client builder * Update becaon node binary to use client builder * Ensure release tests compile * Remove old eth1 crate * Add first pass of new lighthouse binary * Fix websocket server startup * Remove old binary code from beacon_node crate * Add first working beacon node tests * Add genesis crate, new eth1 cache_2 * Add Serivce to Eth1Cache * Refactor with general eth1 improvements * Add passing genesis test * Tidy, add comments * Add more comments to eth1 service * Add further eth1 progress * Fix some bugs with genesis * Fix eth1 bugs, make eth1 linking more efficient * Shift logic in genesis service * Add more comments to genesis service * Add gzip, max request values, timeouts to http * Update testnet parameters to suit goerli testnet * Add ability to vary Fork, fix custom spec * Be more explicit about deposit fork version * Start adding beacon chain eth1 option * Add more flexibility to prod client * Further runtime refactoring * Allow for starting from store * Add bootstrapping to client config * Add remote_beacon_node crate * Update eth1 service for more configurability * Update eth1 tests to use less runtimes * Patch issues with tests using too many files * Move dummy eth1 backend flag * Ensure all tests pass * Add ganache-cli to Dockerfile * Use a special docker hub image for testing * Appease clippy * Move validator client into lighthouse binary * Allow starting with dummy eth1 backend * Improve logging * Fix dummy eth1 backend from cli * Add extra testnet command * Ensure consistent spec in beacon node * Update eth1 rig to work on goerli * Tidy lcli, start adding support for yaml config * Add incomplete YamlConfig struct * Remove efforts at YamlConfig * Add incomplete eth1 voting. Blocked on spec issues * Add (untested) first pass at eth1 vote algo * Add tests for winning vote * Add more tests for eth1 chain * Add more eth1 voting tests * Added more eth1 voting testing * Change test name * Add more tests to eth1 chain * Tidy eth1 generics, add more tests * Improve comments * Tidy beacon_node tests * Tidy, rename JsonRpc.. to Caching.. * Tidy voting logic * Tidy builder docs * Add comments, tidy eth1 * Add more comments to eth1 * Fix bug with winning_vote * Add doc comments to the `ClientBuilder` * Remove commented-out code * Improve `ClientBuilder` docs * Add comments to client config * Add decoding test for `ClientConfig` * Remove unused `DepositSet` struct * Tidy `block_cache` * Remove commented out lines * Remove unused code in `eth1` crate * Remove old validator binary `main.rs` * Tidy, fix tests compile error * Add initial tests for get_deposits * Remove dead code in eth1_test_rig * Update TestingDepositBuilder * Add testing for getting eth1 deposits * Fix duplicate rand dep * Remove dead code * Remove accidentally-added files * Fix comment in eth1_genesis_service * Add .gitignore for eth1_test_rig * Fix bug in eth1_genesis_service * Remove dead code from eth2_config * Fix tabs/spaces in root Cargo.toml * Tidy eth1 crate * Allow for re-use of eth1 service after genesis * Update docs for new CLI * Change README gif * Tidy eth1 http module * Tidy eth1 service * Tidy environment crate * Remove unused file * Tidy, add comments * Remove commented-out code * Address majority of Michael's comments * Address other PR comments * Add link to issue alongside TODO
2019-11-15 03:47:51 +00:00
builder::{BeaconChainBuilder, Witness},
eth1_chain::CachingEth1Backend,
BeaconChain, BeaconChainTypes, BlockError, ChainConfig, ServerSentEventHandler,
StateSkipConfig,
};
use bls::get_withdrawal_credentials;
use execution_layer::test_utils::DEFAULT_JWT_SECRET;
use execution_layer::{
auth::JwtKey,
test_utils::{ExecutionBlockGenerator, MockExecutionLayer, DEFAULT_TERMINAL_BLOCK},
ExecutionLayer,
};
use futures::channel::mpsc::Receiver;
pub use genesis::{interop_genesis_state, DEFAULT_ETH1_BLOCK_HASH};
use int_to_bytes::int_to_bytes32;
use merkle_proof::MerkleTree;
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 parking_lot::Mutex;
use parking_lot::RwLockWriteGuard;
use rand::rngs::StdRng;
use rand::Rng;
use rand::SeedableRng;
use rayon::prelude::*;
use sensitive_url::SensitiveUrl;
use slog::Logger;
2019-06-16 06:24:33 +00:00
use slot_clock::TestingSlotClock;
use state_processing::per_block_processing::compute_timestamp_at_slot;
use state_processing::{
state_advance::{complete_state_advance, partial_state_advance},
StateRootStrategy,
};
Faster attestation production (#838) * Start adding interop genesis state to lcli * Use more efficient method to generate genesis state * Remove duplicate int_to_bytes32 * Add lcli command to change state genesis time * Add option to allow VC to start with unsynced BN * Set VC to do parallel key loading * Don't default to dummy eth1 backend * Add endpoint to dump operation pool * Add metrics for op pool * Remove state clone for slot notifier * Add mem size approximation for tree hash cache * Avoid cloning tree hash when getting head * Avoid cloning tree hash when getting head * Add working arena-based cached tree hash * Add another benchmark * Add pre-allocation for caches * Make cache nullable * Fix bugs in cache tree hash * Add validator tree hash optimization * Optimize hash_concat * Make hash32_concat return fixed-len array * Fix failing API tests * Add new beacon state cache struct * Add validator-specific cache * Separate list and values arenas * Add parallel validator registry hashing * Remove MultiTreeHashCache * Remove cached tree hash macro * Fix failing tree hash test * Address Michael's comments * Add CachedTreeHash impl for ef tests * Fix messy merge conflict * Optimize attestation production * Add first basic optimizations * Fix SlotOutOfBounds error * Resolved missed merge conflicts * Fix another missed merge conflict * Fix more merge conflict issues * Add `StateSkipConfig` * Fix test compile errors * Add failing test * Fix bug, make tests pass * Add comment * Delete unused function * Replace deleted comment
2020-03-04 06:10:22 +00:00
use std::borrow::Cow;
use std::collections::{HashMap, HashSet};
Add `execution_optimistic` flag to HTTP responses (#3070) ## Issue Addressed #3031 ## Proposed Changes Updates the following API endpoints to conform with https://github.com/ethereum/beacon-APIs/pull/190 and https://github.com/ethereum/beacon-APIs/pull/196 - [x] `beacon/states/{state_id}/root` - [x] `beacon/states/{state_id}/fork` - [x] `beacon/states/{state_id}/finality_checkpoints` - [x] `beacon/states/{state_id}/validators` - [x] `beacon/states/{state_id}/validators/{validator_id}` - [x] `beacon/states/{state_id}/validator_balances` - [x] `beacon/states/{state_id}/committees` - [x] `beacon/states/{state_id}/sync_committees` - [x] `beacon/headers` - [x] `beacon/headers/{block_id}` - [x] `beacon/blocks/{block_id}` - [x] `beacon/blocks/{block_id}/root` - [x] `beacon/blocks/{block_id}/attestations` - [x] `debug/beacon/states/{state_id}` - [x] `debug/beacon/heads` - [x] `validator/duties/attester/{epoch}` - [x] `validator/duties/proposer/{epoch}` - [x] `validator/duties/sync/{epoch}` Updates the following Server-Sent Events: - [x] `events?topics=head` - [x] `events?topics=block` - [x] `events?topics=finalized_checkpoint` - [x] `events?topics=chain_reorg` ## Backwards Incompatible There is a very minor breaking change with the way the API now handles requests to `beacon/blocks/{block_id}/root` and `beacon/states/{state_id}/root` when `block_id` or `state_id` is the `Root` variant of `BlockId` and `StateId` respectively. Previously a request to a non-existent root would simply echo the root back to the requester: ``` curl "http://localhost:5052/eth/v1/beacon/states/0xaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa/root" {"data":{"root":"0xaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"}} ``` Now it will return a `404`: ``` curl "http://localhost:5052/eth/v1/beacon/blocks/0xaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa/root" {"code":404,"message":"NOT_FOUND: beacon block with root 0xaaaa…aaaa","stacktraces":[]} ``` In addition to this is the block root `0x0000000000000000000000000000000000000000000000000000000000000000` previously would return the genesis block. It will now return a `404`: ``` curl "http://localhost:5052/eth/v1/beacon/blocks/0x0000000000000000000000000000000000000000000000000000000000000000" {"code":404,"message":"NOT_FOUND: beacon block with root 0x0000…0000","stacktraces":[]} ``` ## Additional Info - `execution_optimistic` is always set, and will return `false` pre-Bellatrix. I am also open to the idea of doing something like `#[serde(skip_serializing_if = "Option::is_none")]`. - The value of `execution_optimistic` is set to `false` where possible. Any computation that is reliant on the `head` will simply use the `ExecutionStatus` of the head (unless the head block is pre-Bellatrix). Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-07-25 08:23:00 +00:00
use std::fmt;
use std::str::FromStr;
2019-06-16 06:24:33 +00:00
use std::sync::Arc;
Eth1 Integration (#542) * Refactor to cache Eth1Data * Fix merge conflicts and minor refactorings * Rename Eth1Cache to Eth1DataCache * Refactor events subscription * Add deposits module to interface with BeaconChain deposits * Remove utils * Rename to types.rs and add trait constraints to Eth1DataFetcher * Confirm to trait constraints. Make Web3DataFetcher cloneable * Make fetcher object member of deposit and eth1_data cache and other fixes * Fix update_cache function * Move fetch_eth1_data to impl block * Fix deposit tests * Create Eth1 object for interfacing with Beacon chain * Add `run` function for running update_cache and subscribe_deposit_logs tasks * Add logging * Run `cargo fmt` and make tests pass * Convert sync functions to async * Add timeouts to web3 functions * Return futures from cache functions * Add failed chaining of futures * Working cache updation * Clean up tests and `update_cache` function * Refactor `get_eth1_data` functions to work with future returning functions * Refactor eth1 `run` function to work with modified `update_cache` api * Minor changes * Add distance parameter to `update_cache` * Fix tests and other minor fixes * Working integration with cache and deposits * Add merkle_tree construction, proof generation and verification code * Add function to construct and fetch Deposits for BeaconNode * Add error handling * Import ssz * Add error handling to eth1 cache and fix minor errors * Run rustfmt * Fix minor bug * Rename Eth1Error and change to Result<T> * Change deposit fetching mechanism from notification based to poll based * Add deposits from eth1 chain in a given range every `x` blocks * Modify `run` function to accommodate changes * Minor fixes * Fix formatting * Initial commit. web3 api working. * Tidied up lib. Add function for fetching logs. * Refactor with `Eth1DataFetcher` trait * Add parsing for deposit contract logs and get_eth1_data function * Add `get_eth1_votes` function * Refactor to cache Eth1Data * Fix merge conflicts and minor refactorings * Rename Eth1Cache to Eth1DataCache * Refactor events subscription * Add deposits module to interface with BeaconChain deposits * Remove utils * Rename to types.rs and add trait constraints to Eth1DataFetcher * Confirm to trait constraints. Make Web3DataFetcher cloneable * Make fetcher object member of deposit and eth1_data cache and other fixes * Fix update_cache function * Move fetch_eth1_data to impl block * Fix deposit tests * Create Eth1 object for interfacing with Beacon chain * Add `run` function for running update_cache and subscribe_deposit_logs tasks * Add logging * Run `cargo fmt` and make tests pass * Convert sync functions to async * Add timeouts to web3 functions * Return futures from cache functions * Add failed chaining of futures * Working cache updation * Clean up tests and `update_cache` function * Refactor `get_eth1_data` functions to work with future returning functions * Refactor eth1 `run` function to work with modified `update_cache` api * Minor changes * Add distance parameter to `update_cache` * Fix tests and other minor fixes * Working integration with cache and deposits * Add merkle_tree construction, proof generation and verification code * Add function to construct and fetch Deposits for BeaconNode * Add error handling * Import ssz * Add error handling to eth1 cache and fix minor errors * Run rustfmt * Fix minor bug * Rename Eth1Error and change to Result<T> * Change deposit fetching mechanism from notification based to poll based * Add deposits from eth1 chain in a given range every `x` blocks * Modify `run` function to accommodate changes * Minor fixes * Fix formatting * Fix merge issue * Refactor with `Config` struct. Remote `ContractConfig` * Rename eth1_chain crate to eth1 * Rename files and read abi file using `fs::read` * Move eth1 to lib * Remove unnecessary mutability constraint * Add `Web3Backend` for returning actual eth1 data * Refactor `get_eth1_votes` to return a Result * Delete `eth1_chain` crate * Return `Result` from `get_deposits` * Fix range of deposits to return to beacon chain * Add `get_block_height_by_hash` trait function * Add naive method for getting `previous_eth1_distance` * Add eth1 config params to main config * Add instructions for setting up eth1 testing environment * Add build script to fetch deposit contract abi * Contract ABI is part of compiled binary * Fix minor bugs * Move docs to lib * Add timeout to config * Remove print statements * Change warn to error * Fix typos * Removed prints in test and get timeout value from config * Fixed error types * Added logging to web3_fetcher * Refactor for modified web3 api * Fix minor stuff * Add build script * Tidy, hide eth1 integration tests behind flag * Add http crate * Add first stages of eth1_test_rig * Fix deposits on test rig * Fix bug with deposit count method * Add block hash getter to http eth1 * Clean eth1 http crate and tests * Add script to start ganache * Adds deposit tree to eth1-http * Extend deposit tree tests * Tidy tests in eth1-http * Add more detail to get block request * Add block cache to eth1-http * Rename deposit tree to deposit cache * Add inital updating to eth1-http * Tidy updater * Fix compile bugs in tests * Adds an Eth1DataCache builder * Reorg eth1-http files * Add (failing) tests for eth1 updater * Rename files, fix bug in eth1-http * Ensure that ganache timestamps are increasing * Fix bugs with getting eth1data ancestors * Improve eth1 testing, fix bugs * Add truncate method to block cache * Add pruning to block cache update process * Add tests for block pruning * Allow for dropping an expired cache. * Add more comments * Add first compiling version of deposit updater * Add common fn for getting range of required blocks * Add passing deposit update test * Improve tests * Fix block pruning bug * Add tests for running two updates at once * Add updater services to eth1 * Add deposit collection to beacon chain * Add incomplete builder experiments * Add first working version of beacon chain builder * Update test harness to new beacon chain type * Rename builder file, tidy * Add first working client builder * Progress further on client builder * Update becaon node binary to use client builder * Ensure release tests compile * Remove old eth1 crate * Add first pass of new lighthouse binary * Fix websocket server startup * Remove old binary code from beacon_node crate * Add first working beacon node tests * Add genesis crate, new eth1 cache_2 * Add Serivce to Eth1Cache * Refactor with general eth1 improvements * Add passing genesis test * Tidy, add comments * Add more comments to eth1 service * Add further eth1 progress * Fix some bugs with genesis * Fix eth1 bugs, make eth1 linking more efficient * Shift logic in genesis service * Add more comments to genesis service * Add gzip, max request values, timeouts to http * Update testnet parameters to suit goerli testnet * Add ability to vary Fork, fix custom spec * Be more explicit about deposit fork version * Start adding beacon chain eth1 option * Add more flexibility to prod client * Further runtime refactoring * Allow for starting from store * Add bootstrapping to client config * Add remote_beacon_node crate * Update eth1 service for more configurability * Update eth1 tests to use less runtimes * Patch issues with tests using too many files * Move dummy eth1 backend flag * Ensure all tests pass * Add ganache-cli to Dockerfile * Use a special docker hub image for testing * Appease clippy * Move validator client into lighthouse binary * Allow starting with dummy eth1 backend * Improve logging * Fix dummy eth1 backend from cli * Add extra testnet command * Ensure consistent spec in beacon node * Update eth1 rig to work on goerli * Tidy lcli, start adding support for yaml config * Add incomplete YamlConfig struct * Remove efforts at YamlConfig * Add incomplete eth1 voting. Blocked on spec issues * Add (untested) first pass at eth1 vote algo * Add tests for winning vote * Add more tests for eth1 chain * Add more eth1 voting tests * Added more eth1 voting testing * Change test name * Add more tests to eth1 chain * Tidy eth1 generics, add more tests * Improve comments * Tidy beacon_node tests * Tidy, rename JsonRpc.. to Caching.. * Tidy voting logic * Tidy builder docs * Add comments, tidy eth1 * Add more comments to eth1 * Fix bug with winning_vote * Add doc comments to the `ClientBuilder` * Remove commented-out code * Improve `ClientBuilder` docs * Add comments to client config * Add decoding test for `ClientConfig` * Remove unused `DepositSet` struct * Tidy `block_cache` * Remove commented out lines * Remove unused code in `eth1` crate * Remove old validator binary `main.rs` * Tidy, fix tests compile error * Add initial tests for get_deposits * Remove dead code in eth1_test_rig * Update TestingDepositBuilder * Add testing for getting eth1 deposits * Fix duplicate rand dep * Remove dead code * Remove accidentally-added files * Fix comment in eth1_genesis_service * Add .gitignore for eth1_test_rig * Fix bug in eth1_genesis_service * Remove dead code from eth2_config * Fix tabs/spaces in root Cargo.toml * Tidy eth1 crate * Allow for re-use of eth1 service after genesis * Update docs for new CLI * Change README gif * Tidy eth1 http module * Tidy eth1 service * Tidy environment crate * Remove unused file * Tidy, add comments * Remove commented-out code * Address majority of Michael's comments * Address other PR comments * Add link to issue alongside TODO
2019-11-15 03:47:51 +00:00
use std::time::Duration;
use store::{config::StoreConfig, HotColdDB, ItemStore, LevelDB, MemoryStore};
use task_executor::{test_utils::TestRuntime, ShutdownReason};
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 tree_hash::TreeHash;
use types::sync_selection_proof::SyncSelectionProof;
pub use types::test_utils::generate_deterministic_keypairs;
use types::{typenum::U4294967296, *};
2019-06-16 06:24:33 +00:00
Prepare for public testnet (#628) * Update to spec v0.9.0 * Update to v0.9.1 * Bump spec tags for v0.9.1 * Formatting, fix CI failures * Resolve accidental KeyPair merge conflict * Document new BeaconState functions * Add `validator` changes from `validator-to-rest` * Add initial (failing) REST api tests * Fix signature parsing * Add more tests * Refactor http router * Add working tests for publish beacon block * Add validator duties tests * Move account_manager under `lighthouse` binary * Unify logfile handling in `environment` crate. * Fix incorrect cache drops in `advance_caches` * Update fork choice for v0.9.1 * Add `deposit_contract` crate * Add progress on validator onboarding * Add unfinished attesation code * Update account manager CLI * Write eth1 data file as hex string * Integrate ValidatorDirectory with validator_client * Move ValidatorDirectory into validator_client * Clean up some FIXMEs * Add beacon_chain_sim * Fix a few docs/logs * Expand `beacon_chain_sim` * Fix spec for `beacon_chain_sim * More testing for api * Start work on attestation endpoint * Reject empty attestations * Allow attestations to genesis block * Add working tests for `rest_api` validator endpoint * Remove grpc from beacon_node * Start heavy refactor of validator client - Block production is working * Prune old validator client files * Start works on attestation service * Add attestation service to validator client * Use full pubkey for validator directories * Add validator duties post endpoint * Use par_iter for keypair generation * Use bulk duties request in validator client * Add version http endpoint tests * Add interop keys and startup wait * Ensure a prompt exit * Add duties pruning * Fix compile error in beacon node tests * Add github workflow * Modify rust.yaml * Modify gitlab actions * Add to CI file * Add sudo to CI npm install * Move cargo fmt to own job in tests * Fix cargo fmt in CI * Add rustup update before cargo fmt * Change name of CI job * Make other CI jobs require cargo fmt * Add CI badge * Remove gitlab and travis files * Add different http timeout for debug * Update docker file, use makefile in CI * Use make in the dockerfile, skip the test * Use the makefile for debug GI test * Update book * Tidy grpc and misc things * Apply discv5 fixes * Address other minor issues * Fix warnings * Attempt fix for addr parsing * Tidy validator config, CLIs * Tidy comments * Tidy signing, reduce ForkService duplication * Fail if skipping too many slots * Set default recent genesis time to 0 * Add custom http timeout to validator * Fix compile bug in node_test_rig * Remove old bootstrap flag from val CLI * Update docs * Tidy val client * Change val client log levels * Add comments, more validity checks * Fix compile error, add comments * Undo changes to eth2-libp2p/src * Reduce duplication of keypair generation * Add more logging for validator duties * Fix beacon_chain_sim, nitpicks * Fix compile error, minor nits * Update to use v0.9.2 version of deposit contract * Add efforts to automate eth1 testnet deployment * Fix lcli testnet deployer * Modify bn CLI to parse eth2_testnet_dir * Progress with account_manager deposit tools * Make account manager submit deposits * Add password option for submitting deposits * Allow custom deposit amount * Add long names to lcli clap * Add password option to lcli deploy command * Fix minor bugs whilst testing * Address Michael's comments * Add refund-deposit-contract to lcli * Use time instead of skip count for denying long skips * Improve logging for eth1 * Fix bug with validator services exiting on error * Drop the block cache after genesis * Modify eth1 testnet config * Improve eth1 logging * Make validator wait until genesis time * Fix bug in eth1 voting * Add more logging to eth1 voting * Handle errors in eth1 http module * Set SECONDS_PER_DAY to sensible minimum * Shorten delay before testnet start * Ensure eth1 block is produced without any votes * Improve eth1 logging * Fix broken tests in eth1 * Tidy code in rest_api * Fix failing test in deposit_contract * Make CLI args more consistent * Change validator/duties endpoint * Add time-based skip slot limiting * Add new error type missed in previous commit * Add log when waiting for genesis * Refactor beacon node CLI * Remove unused dep * Add lcli eth1-genesis command * Fix bug in master merge * Apply clippy lints to beacon node * Add support for YamlConfig in Eth2TestnetDir * Upgrade tesnet deposit contract version * Remove unnecessary logging and correct formatting * Add a hardcoded eth2 testnet config * Ensure http server flag works. Overwrite configs with flags. * Ensure boot nodes are loaded from testnet dir * Fix account manager CLI bugs * Fix bugs with beacon node cli * Allow testnet dir without boot nodes * Write genesis state as SSZ * Remove ---/n from the start of testnet_dir files * Set default libp2p address * Tidy account manager CLI, add logging * Add check to see if testnet dir exists * Apply reviewers suggestions * Add HeadTracker struct * Add fork choice persistence * Shorten slot time for simulator * Add the /beacon/heads API endpoint * Update hardcoded testnet * Add tests for BeaconChain persistence + fix bugs * Extend BeaconChain persistence testing * Ensure chain is finalized b4 persistence tests * Ensure boot_enr.yaml is include in binary * Refactor beacon_chain_sim * Move files about in beacon sim * Update beacon_chain_sim * Fix bug with deposit inclusion * Increase log in genesis service, fix todo * Tidy sim, fix broken rest_api tests * Fix more broken tests * Update testnet * Fix broken rest api test * Tidy account manager CLI * Use tempdir for account manager * Stop hardcoded testnet dir from creating dir * Rename Eth2TestnetDir to Eth2TestnetConfig * Change hardcoded -> hard_coded * Tidy account manager * Add log to account manager * Tidy, ensure head tracker is loaded from disk * Tidy beacon chain builder * Tidy eth1_chain * Adds log support for simulator * Revert "Adds log support for simulator" This reverts commit ec77c66a052350f551db145cf20f213823428dd3. * Adds log support for simulator * Tidy after self-review * Change default log level * Address Michael's delicious PR comments * Fix off-by-one in tests
2019-12-03 04:28:57 +00:00
// 4th September 2019
pub const HARNESS_GENESIS_TIME: u64 = 1_567_552_690;
// Environment variable to read if `fork_from_env` feature is enabled.
const FORK_NAME_ENV_VAR: &str = "FORK_NAME";
// Default target aggregators to set during testing, this ensures an aggregator at each slot.
//
// You should mutate the `ChainSpec` prior to initialising the harness if you would like to use
// a different value.
pub const DEFAULT_TARGET_AGGREGATORS: u64 = u64::max_value();
pub type BaseHarnessType<TEthSpec, THotStore, TColdStore> =
Witness<TestingSlotClock, CachingEth1Backend<TEthSpec>, TEthSpec, THotStore, TColdStore>;
Eth1 Integration (#542) * Refactor to cache Eth1Data * Fix merge conflicts and minor refactorings * Rename Eth1Cache to Eth1DataCache * Refactor events subscription * Add deposits module to interface with BeaconChain deposits * Remove utils * Rename to types.rs and add trait constraints to Eth1DataFetcher * Confirm to trait constraints. Make Web3DataFetcher cloneable * Make fetcher object member of deposit and eth1_data cache and other fixes * Fix update_cache function * Move fetch_eth1_data to impl block * Fix deposit tests * Create Eth1 object for interfacing with Beacon chain * Add `run` function for running update_cache and subscribe_deposit_logs tasks * Add logging * Run `cargo fmt` and make tests pass * Convert sync functions to async * Add timeouts to web3 functions * Return futures from cache functions * Add failed chaining of futures * Working cache updation * Clean up tests and `update_cache` function * Refactor `get_eth1_data` functions to work with future returning functions * Refactor eth1 `run` function to work with modified `update_cache` api * Minor changes * Add distance parameter to `update_cache` * Fix tests and other minor fixes * Working integration with cache and deposits * Add merkle_tree construction, proof generation and verification code * Add function to construct and fetch Deposits for BeaconNode * Add error handling * Import ssz * Add error handling to eth1 cache and fix minor errors * Run rustfmt * Fix minor bug * Rename Eth1Error and change to Result<T> * Change deposit fetching mechanism from notification based to poll based * Add deposits from eth1 chain in a given range every `x` blocks * Modify `run` function to accommodate changes * Minor fixes * Fix formatting * Initial commit. web3 api working. * Tidied up lib. Add function for fetching logs. * Refactor with `Eth1DataFetcher` trait * Add parsing for deposit contract logs and get_eth1_data function * Add `get_eth1_votes` function * Refactor to cache Eth1Data * Fix merge conflicts and minor refactorings * Rename Eth1Cache to Eth1DataCache * Refactor events subscription * Add deposits module to interface with BeaconChain deposits * Remove utils * Rename to types.rs and add trait constraints to Eth1DataFetcher * Confirm to trait constraints. Make Web3DataFetcher cloneable * Make fetcher object member of deposit and eth1_data cache and other fixes * Fix update_cache function * Move fetch_eth1_data to impl block * Fix deposit tests * Create Eth1 object for interfacing with Beacon chain * Add `run` function for running update_cache and subscribe_deposit_logs tasks * Add logging * Run `cargo fmt` and make tests pass * Convert sync functions to async * Add timeouts to web3 functions * Return futures from cache functions * Add failed chaining of futures * Working cache updation * Clean up tests and `update_cache` function * Refactor `get_eth1_data` functions to work with future returning functions * Refactor eth1 `run` function to work with modified `update_cache` api * Minor changes * Add distance parameter to `update_cache` * Fix tests and other minor fixes * Working integration with cache and deposits * Add merkle_tree construction, proof generation and verification code * Add function to construct and fetch Deposits for BeaconNode * Add error handling * Import ssz * Add error handling to eth1 cache and fix minor errors * Run rustfmt * Fix minor bug * Rename Eth1Error and change to Result<T> * Change deposit fetching mechanism from notification based to poll based * Add deposits from eth1 chain in a given range every `x` blocks * Modify `run` function to accommodate changes * Minor fixes * Fix formatting * Fix merge issue * Refactor with `Config` struct. Remote `ContractConfig` * Rename eth1_chain crate to eth1 * Rename files and read abi file using `fs::read` * Move eth1 to lib * Remove unnecessary mutability constraint * Add `Web3Backend` for returning actual eth1 data * Refactor `get_eth1_votes` to return a Result * Delete `eth1_chain` crate * Return `Result` from `get_deposits` * Fix range of deposits to return to beacon chain * Add `get_block_height_by_hash` trait function * Add naive method for getting `previous_eth1_distance` * Add eth1 config params to main config * Add instructions for setting up eth1 testing environment * Add build script to fetch deposit contract abi * Contract ABI is part of compiled binary * Fix minor bugs * Move docs to lib * Add timeout to config * Remove print statements * Change warn to error * Fix typos * Removed prints in test and get timeout value from config * Fixed error types * Added logging to web3_fetcher * Refactor for modified web3 api * Fix minor stuff * Add build script * Tidy, hide eth1 integration tests behind flag * Add http crate * Add first stages of eth1_test_rig * Fix deposits on test rig * Fix bug with deposit count method * Add block hash getter to http eth1 * Clean eth1 http crate and tests * Add script to start ganache * Adds deposit tree to eth1-http * Extend deposit tree tests * Tidy tests in eth1-http * Add more detail to get block request * Add block cache to eth1-http * Rename deposit tree to deposit cache * Add inital updating to eth1-http * Tidy updater * Fix compile bugs in tests * Adds an Eth1DataCache builder * Reorg eth1-http files * Add (failing) tests for eth1 updater * Rename files, fix bug in eth1-http * Ensure that ganache timestamps are increasing * Fix bugs with getting eth1data ancestors * Improve eth1 testing, fix bugs * Add truncate method to block cache * Add pruning to block cache update process * Add tests for block pruning * Allow for dropping an expired cache. * Add more comments * Add first compiling version of deposit updater * Add common fn for getting range of required blocks * Add passing deposit update test * Improve tests * Fix block pruning bug * Add tests for running two updates at once * Add updater services to eth1 * Add deposit collection to beacon chain * Add incomplete builder experiments * Add first working version of beacon chain builder * Update test harness to new beacon chain type * Rename builder file, tidy * Add first working client builder * Progress further on client builder * Update becaon node binary to use client builder * Ensure release tests compile * Remove old eth1 crate * Add first pass of new lighthouse binary * Fix websocket server startup * Remove old binary code from beacon_node crate * Add first working beacon node tests * Add genesis crate, new eth1 cache_2 * Add Serivce to Eth1Cache * Refactor with general eth1 improvements * Add passing genesis test * Tidy, add comments * Add more comments to eth1 service * Add further eth1 progress * Fix some bugs with genesis * Fix eth1 bugs, make eth1 linking more efficient * Shift logic in genesis service * Add more comments to genesis service * Add gzip, max request values, timeouts to http * Update testnet parameters to suit goerli testnet * Add ability to vary Fork, fix custom spec * Be more explicit about deposit fork version * Start adding beacon chain eth1 option * Add more flexibility to prod client * Further runtime refactoring * Allow for starting from store * Add bootstrapping to client config * Add remote_beacon_node crate * Update eth1 service for more configurability * Update eth1 tests to use less runtimes * Patch issues with tests using too many files * Move dummy eth1 backend flag * Ensure all tests pass * Add ganache-cli to Dockerfile * Use a special docker hub image for testing * Appease clippy * Move validator client into lighthouse binary * Allow starting with dummy eth1 backend * Improve logging * Fix dummy eth1 backend from cli * Add extra testnet command * Ensure consistent spec in beacon node * Update eth1 rig to work on goerli * Tidy lcli, start adding support for yaml config * Add incomplete YamlConfig struct * Remove efforts at YamlConfig * Add incomplete eth1 voting. Blocked on spec issues * Add (untested) first pass at eth1 vote algo * Add tests for winning vote * Add more tests for eth1 chain * Add more eth1 voting tests * Added more eth1 voting testing * Change test name * Add more tests to eth1 chain * Tidy eth1 generics, add more tests * Improve comments * Tidy beacon_node tests * Tidy, rename JsonRpc.. to Caching.. * Tidy voting logic * Tidy builder docs * Add comments, tidy eth1 * Add more comments to eth1 * Fix bug with winning_vote * Add doc comments to the `ClientBuilder` * Remove commented-out code * Improve `ClientBuilder` docs * Add comments to client config * Add decoding test for `ClientConfig` * Remove unused `DepositSet` struct * Tidy `block_cache` * Remove commented out lines * Remove unused code in `eth1` crate * Remove old validator binary `main.rs` * Tidy, fix tests compile error * Add initial tests for get_deposits * Remove dead code in eth1_test_rig * Update TestingDepositBuilder * Add testing for getting eth1 deposits * Fix duplicate rand dep * Remove dead code * Remove accidentally-added files * Fix comment in eth1_genesis_service * Add .gitignore for eth1_test_rig * Fix bug in eth1_genesis_service * Remove dead code from eth2_config * Fix tabs/spaces in root Cargo.toml * Tidy eth1 crate * Allow for re-use of eth1 service after genesis * Update docs for new CLI * Change README gif * Tidy eth1 http module * Tidy eth1 service * Tidy environment crate * Remove unused file * Tidy, add comments * Remove commented-out code * Address majority of Michael's comments * Address other PR comments * Add link to issue alongside TODO
2019-11-15 03:47:51 +00:00
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
pub type DiskHarnessType<E> = BaseHarnessType<E, LevelDB<E>, LevelDB<E>>;
pub type EphemeralHarnessType<E> = BaseHarnessType<E, MemoryStore<E>, MemoryStore<E>>;
Run fork choice before block proposal (#3168) ## Issue Addressed Upcoming spec change https://github.com/ethereum/consensus-specs/pull/2878 ## Proposed Changes 1. Run fork choice at the start of every slot, and wait for this run to complete before proposing a block. 2. As an optimisation, also run fork choice 3/4 of the way through the slot (at 9s), _dequeueing attestations for the next slot_. 3. Remove the fork choice run from the state advance timer that occurred before advancing the state. ## Additional Info ### Block Proposal Accuracy This change makes us more likely to propose on top of the correct head in the presence of re-orgs with proposer boost in play. The main scenario that this change is designed to address is described in the linked spec issue. ### Attestation Accuracy This change _also_ makes us more likely to attest to the correct head. Currently in the case of a skipped slot at `slot` we only run fork choice 9s into `slot - 1`. This means the attestations from `slot - 1` aren't taken into consideration, and any boost applied to the block from `slot - 1` is not removed (it should be). In the language of the linked spec issue, this means we are liable to attest to C, even when the majority voting weight has already caused a re-org to B. ### Why remove the call before the state advance? If we've run fork choice at the start of the slot then it has already dequeued all the attestations from the previous slot, which are the only ones eligible to influence the head in the current slot. Running fork choice again is unnecessary (unless we run it for the next slot and try to pre-empt a re-org, but I don't currently think this is a great idea). ### Performance Based on Prater testing this adds about 5-25ms of runtime to block proposal times, which are 500-1000ms on average (and spike to 5s+ sometimes due to state handling issues :cry: ). I believe this is a small enough penalty to enable it by default, with the option to disable it via the new flag `--fork-choice-before-proposal-timeout 0`. Upcoming work on block packing and state representation will also reduce block production times in general, while removing the spikes. ### Implementation Fork choice gets invoked at the start of the slot via the `per_slot_task` function called from the slot timer. It then uses a condition variable to signal to block production that fork choice has been updated. This is a bit funky, but it seems to work. One downside of the timer-based approach is that it doesn't happen automatically in most of the tests. The test added by this PR has to trigger the run manually.
2022-05-20 05:02:11 +00:00
pub type BoxedMutator<E, Hot, Cold> = Box<
dyn FnOnce(
BeaconChainBuilder<BaseHarnessType<E, Hot, Cold>>,
) -> BeaconChainBuilder<BaseHarnessType<E, Hot, Cold>>,
>;
pub type AddBlocksResult<E> = (
HashMap<Slot, SignedBeaconBlockHash>,
HashMap<Slot, BeaconStateHash>,
SignedBeaconBlockHash,
BeaconState<E>,
);
/// Deprecated: Indicates how the `BeaconChainHarness` should produce blocks.
2019-06-20 00:59:19 +00:00
#[derive(Clone, Copy, Debug)]
2019-06-21 01:55:37 +00:00
pub enum BlockStrategy {
/// Produce blocks upon the canonical head (normal case).
OnCanonicalHead,
2019-06-21 01:55:37 +00:00
/// Ignore the canonical head and produce blocks upon the block at the given slot.
///
/// Useful for simulating forks.
2019-06-23 00:05:12 +00:00
ForkCanonicalChainAt {
/// The slot of the parent of the first block produced.
previous_slot: Slot,
/// The slot of the first block produced (must be higher than `previous_slot`.
first_slot: Slot,
},
}
/// Deprecated: Indicates how the `BeaconChainHarness` should produce attestations.
2019-06-21 08:54:37 +00:00
#[derive(Clone, Debug)]
pub enum AttestationStrategy {
2019-06-21 01:55:37 +00:00
/// All validators attest to whichever block the `BeaconChainHarness` has produced.
AllValidators,
2019-06-21 08:54:37 +00:00
/// Only the given validators should attest. All others should fail to produce attestations.
SomeValidators(Vec<usize>),
}
/// Indicates whether the `BeaconChainHarness` should use the `state.current_sync_committee` or
/// `state.next_sync_committee` when creating sync messages or contributions.
#[derive(Clone, Debug)]
pub enum RelativeSyncCommittee {
Current,
Next,
}
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
fn make_rng() -> Mutex<StdRng> {
// Nondeterminism in tests is a highly undesirable thing. Seed the RNG to some arbitrary
// but fixed value for reproducibility.
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
Mutex::new(StdRng::seed_from_u64(0x0DDB1A5E5BAD5EEDu64))
}
/// Return a `ChainSpec` suitable for test usage.
///
/// If the `fork_from_env` feature is enabled, read the fork to use from the FORK_NAME environment
/// variable. Otherwise use the default spec.
pub fn test_spec<E: EthSpec>() -> ChainSpec {
let mut spec = if cfg!(feature = "fork_from_env") {
let fork_name = std::env::var(FORK_NAME_ENV_VAR).unwrap_or_else(|e| {
panic!(
"{} env var must be defined when using fork_from_env: {:?}",
FORK_NAME_ENV_VAR, e
)
});
let fork = ForkName::from_str(fork_name.as_str())
.unwrap_or_else(|()| panic!("unknown FORK_NAME: {}", fork_name));
fork.make_genesis_spec(E::default_spec())
} else {
E::default_spec()
};
// Set target aggregators to a high value by default.
spec.target_aggregators_per_committee = DEFAULT_TARGET_AGGREGATORS;
spec
}
pub struct Builder<T: BeaconChainTypes> {
eth_spec_instance: T::EthSpec,
spec: Option<ChainSpec>,
validator_keypairs: Option<Vec<Keypair>>,
chain_config: Option<ChainConfig>,
store_config: Option<StoreConfig>,
#[allow(clippy::type_complexity)]
store: Option<Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>>,
initial_mutator: Option<BoxedMutator<T::EthSpec, T::HotStore, T::ColdStore>>,
store_mutator: Option<BoxedMutator<T::EthSpec, T::HotStore, T::ColdStore>>,
execution_layer: Option<ExecutionLayer<T::EthSpec>>,
mock_execution_layer: Option<MockExecutionLayer<T::EthSpec>>,
runtime: TestRuntime,
log: Logger,
}
impl<E: EthSpec> Builder<EphemeralHarnessType<E>> {
pub fn fresh_ephemeral_store(mut self) -> Self {
let spec = self.spec.as_ref().expect("cannot build without spec");
let validator_keypairs = self
.validator_keypairs
.clone()
.expect("cannot build without validator keypairs");
let store = Arc::new(
HotColdDB::open_ephemeral(
self.store_config.clone().unwrap_or_default(),
spec.clone(),
self.log.clone(),
)
.unwrap(),
);
let mutator = move |builder: BeaconChainBuilder<_>| {
let genesis_state = interop_genesis_state::<E>(
&validator_keypairs,
HARNESS_GENESIS_TIME,
Hash256::from_slice(DEFAULT_ETH1_BLOCK_HASH),
None,
builder.get_spec(),
)
.expect("should generate interop state");
builder
.genesis_state(genesis_state)
.expect("should build state using recent genesis")
};
self.store = Some(store);
self.store_mutator(Box::new(mutator))
}
/// Create a new ephemeral store that uses the specified `genesis_state`.
pub fn genesis_state_ephemeral_store(mut self, genesis_state: BeaconState<E>) -> Self {
let spec = self.spec.as_ref().expect("cannot build without spec");
let store = Arc::new(
HotColdDB::open_ephemeral(
self.store_config.clone().unwrap_or_default(),
spec.clone(),
self.log.clone(),
)
.unwrap(),
);
let mutator = move |builder: BeaconChainBuilder<_>| {
builder
.genesis_state(genesis_state)
.expect("should build state using recent genesis")
};
self.store = Some(store);
self.store_mutator(Box::new(mutator))
}
}
impl<E: EthSpec> Builder<DiskHarnessType<E>> {
/// Disk store, start from genesis.
pub fn fresh_disk_store(mut self, store: Arc<HotColdDB<E, LevelDB<E>, LevelDB<E>>>) -> Self {
let validator_keypairs = self
.validator_keypairs
.clone()
.expect("cannot build without validator keypairs");
let mutator = move |builder: BeaconChainBuilder<_>| {
let genesis_state = interop_genesis_state::<E>(
&validator_keypairs,
HARNESS_GENESIS_TIME,
Hash256::from_slice(DEFAULT_ETH1_BLOCK_HASH),
None,
builder.get_spec(),
)
.expect("should generate interop state");
builder
.genesis_state(genesis_state)
.expect("should build state using recent genesis")
};
self.store = Some(store);
self.store_mutator(Box::new(mutator))
}
/// Disk store, resume.
pub fn resumed_disk_store(mut self, store: Arc<HotColdDB<E, LevelDB<E>, LevelDB<E>>>) -> Self {
let mutator = move |builder: BeaconChainBuilder<_>| {
builder
.resume_from_db()
.expect("should resume from database")
};
self.store = Some(store);
self.store_mutator(Box::new(mutator))
}
}
impl<E, Hot, Cold> Builder<BaseHarnessType<E, Hot, Cold>>
where
E: EthSpec,
Hot: ItemStore<E>,
Cold: ItemStore<E>,
{
pub fn new(eth_spec_instance: E) -> Self {
let runtime = TestRuntime::default();
let log = runtime.log.clone();
Self {
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
eth_spec_instance,
spec: None,
validator_keypairs: None,
chain_config: None,
store_config: None,
store: None,
initial_mutator: None,
store_mutator: None,
execution_layer: None,
mock_execution_layer: None,
runtime,
log,
}
}
2019-06-16 06:24:33 +00:00
pub fn deterministic_keypairs(self, num_keypairs: usize) -> Self {
self.keypairs(types::test_utils::generate_deterministic_keypairs(
num_keypairs,
))
}
pub fn keypairs(mut self, validator_keypairs: Vec<Keypair>) -> Self {
self.validator_keypairs = Some(validator_keypairs);
self
}
pub fn default_spec(self) -> Self {
self.spec_or_default(None)
2019-06-16 06:24:33 +00:00
}
pub fn spec(self, spec: ChainSpec) -> Self {
self.spec_or_default(Some(spec))
}
pub fn spec_or_default(mut self, spec: Option<ChainSpec>) -> Self {
self.spec = Some(spec.unwrap_or_else(test_spec::<E>));
self
}
Prepare for public testnet (#628) * Update to spec v0.9.0 * Update to v0.9.1 * Bump spec tags for v0.9.1 * Formatting, fix CI failures * Resolve accidental KeyPair merge conflict * Document new BeaconState functions * Add `validator` changes from `validator-to-rest` * Add initial (failing) REST api tests * Fix signature parsing * Add more tests * Refactor http router * Add working tests for publish beacon block * Add validator duties tests * Move account_manager under `lighthouse` binary * Unify logfile handling in `environment` crate. * Fix incorrect cache drops in `advance_caches` * Update fork choice for v0.9.1 * Add `deposit_contract` crate * Add progress on validator onboarding * Add unfinished attesation code * Update account manager CLI * Write eth1 data file as hex string * Integrate ValidatorDirectory with validator_client * Move ValidatorDirectory into validator_client * Clean up some FIXMEs * Add beacon_chain_sim * Fix a few docs/logs * Expand `beacon_chain_sim` * Fix spec for `beacon_chain_sim * More testing for api * Start work on attestation endpoint * Reject empty attestations * Allow attestations to genesis block * Add working tests for `rest_api` validator endpoint * Remove grpc from beacon_node * Start heavy refactor of validator client - Block production is working * Prune old validator client files * Start works on attestation service * Add attestation service to validator client * Use full pubkey for validator directories * Add validator duties post endpoint * Use par_iter for keypair generation * Use bulk duties request in validator client * Add version http endpoint tests * Add interop keys and startup wait * Ensure a prompt exit * Add duties pruning * Fix compile error in beacon node tests * Add github workflow * Modify rust.yaml * Modify gitlab actions * Add to CI file * Add sudo to CI npm install * Move cargo fmt to own job in tests * Fix cargo fmt in CI * Add rustup update before cargo fmt * Change name of CI job * Make other CI jobs require cargo fmt * Add CI badge * Remove gitlab and travis files * Add different http timeout for debug * Update docker file, use makefile in CI * Use make in the dockerfile, skip the test * Use the makefile for debug GI test * Update book * Tidy grpc and misc things * Apply discv5 fixes * Address other minor issues * Fix warnings * Attempt fix for addr parsing * Tidy validator config, CLIs * Tidy comments * Tidy signing, reduce ForkService duplication * Fail if skipping too many slots * Set default recent genesis time to 0 * Add custom http timeout to validator * Fix compile bug in node_test_rig * Remove old bootstrap flag from val CLI * Update docs * Tidy val client * Change val client log levels * Add comments, more validity checks * Fix compile error, add comments * Undo changes to eth2-libp2p/src * Reduce duplication of keypair generation * Add more logging for validator duties * Fix beacon_chain_sim, nitpicks * Fix compile error, minor nits * Update to use v0.9.2 version of deposit contract * Add efforts to automate eth1 testnet deployment * Fix lcli testnet deployer * Modify bn CLI to parse eth2_testnet_dir * Progress with account_manager deposit tools * Make account manager submit deposits * Add password option for submitting deposits * Allow custom deposit amount * Add long names to lcli clap * Add password option to lcli deploy command * Fix minor bugs whilst testing * Address Michael's comments * Add refund-deposit-contract to lcli * Use time instead of skip count for denying long skips * Improve logging for eth1 * Fix bug with validator services exiting on error * Drop the block cache after genesis * Modify eth1 testnet config * Improve eth1 logging * Make validator wait until genesis time * Fix bug in eth1 voting * Add more logging to eth1 voting * Handle errors in eth1 http module * Set SECONDS_PER_DAY to sensible minimum * Shorten delay before testnet start * Ensure eth1 block is produced without any votes * Improve eth1 logging * Fix broken tests in eth1 * Tidy code in rest_api * Fix failing test in deposit_contract * Make CLI args more consistent * Change validator/duties endpoint * Add time-based skip slot limiting * Add new error type missed in previous commit * Add log when waiting for genesis * Refactor beacon node CLI * Remove unused dep * Add lcli eth1-genesis command * Fix bug in master merge * Apply clippy lints to beacon node * Add support for YamlConfig in Eth2TestnetDir * Upgrade tesnet deposit contract version * Remove unnecessary logging and correct formatting * Add a hardcoded eth2 testnet config * Ensure http server flag works. Overwrite configs with flags. * Ensure boot nodes are loaded from testnet dir * Fix account manager CLI bugs * Fix bugs with beacon node cli * Allow testnet dir without boot nodes * Write genesis state as SSZ * Remove ---/n from the start of testnet_dir files * Set default libp2p address * Tidy account manager CLI, add logging * Add check to see if testnet dir exists * Apply reviewers suggestions * Add HeadTracker struct * Add fork choice persistence * Shorten slot time for simulator * Add the /beacon/heads API endpoint * Update hardcoded testnet * Add tests for BeaconChain persistence + fix bugs * Extend BeaconChain persistence testing * Ensure chain is finalized b4 persistence tests * Ensure boot_enr.yaml is include in binary * Refactor beacon_chain_sim * Move files about in beacon sim * Update beacon_chain_sim * Fix bug with deposit inclusion * Increase log in genesis service, fix todo * Tidy sim, fix broken rest_api tests * Fix more broken tests * Update testnet * Fix broken rest api test * Tidy account manager CLI * Use tempdir for account manager * Stop hardcoded testnet dir from creating dir * Rename Eth2TestnetDir to Eth2TestnetConfig * Change hardcoded -> hard_coded * Tidy account manager * Add log to account manager * Tidy, ensure head tracker is loaded from disk * Tidy beacon chain builder * Tidy eth1_chain * Adds log support for simulator * Revert "Adds log support for simulator" This reverts commit ec77c66a052350f551db145cf20f213823428dd3. * Adds log support for simulator * Tidy after self-review * Change default log level * Address Michael's delicious PR comments * Fix off-by-one in tests
2019-12-03 04:28:57 +00:00
/// This mutator will be run before the `store_mutator`.
pub fn initial_mutator(mut self, mutator: BoxedMutator<E, Hot, Cold>) -> Self {
assert!(
self.initial_mutator.is_none(),
"initial mutator already set"
);
self.initial_mutator = Some(mutator);
self
Prepare for public testnet (#628) * Update to spec v0.9.0 * Update to v0.9.1 * Bump spec tags for v0.9.1 * Formatting, fix CI failures * Resolve accidental KeyPair merge conflict * Document new BeaconState functions * Add `validator` changes from `validator-to-rest` * Add initial (failing) REST api tests * Fix signature parsing * Add more tests * Refactor http router * Add working tests for publish beacon block * Add validator duties tests * Move account_manager under `lighthouse` binary * Unify logfile handling in `environment` crate. * Fix incorrect cache drops in `advance_caches` * Update fork choice for v0.9.1 * Add `deposit_contract` crate * Add progress on validator onboarding * Add unfinished attesation code * Update account manager CLI * Write eth1 data file as hex string * Integrate ValidatorDirectory with validator_client * Move ValidatorDirectory into validator_client * Clean up some FIXMEs * Add beacon_chain_sim * Fix a few docs/logs * Expand `beacon_chain_sim` * Fix spec for `beacon_chain_sim * More testing for api * Start work on attestation endpoint * Reject empty attestations * Allow attestations to genesis block * Add working tests for `rest_api` validator endpoint * Remove grpc from beacon_node * Start heavy refactor of validator client - Block production is working * Prune old validator client files * Start works on attestation service * Add attestation service to validator client * Use full pubkey for validator directories * Add validator duties post endpoint * Use par_iter for keypair generation * Use bulk duties request in validator client * Add version http endpoint tests * Add interop keys and startup wait * Ensure a prompt exit * Add duties pruning * Fix compile error in beacon node tests * Add github workflow * Modify rust.yaml * Modify gitlab actions * Add to CI file * Add sudo to CI npm install * Move cargo fmt to own job in tests * Fix cargo fmt in CI * Add rustup update before cargo fmt * Change name of CI job * Make other CI jobs require cargo fmt * Add CI badge * Remove gitlab and travis files * Add different http timeout for debug * Update docker file, use makefile in CI * Use make in the dockerfile, skip the test * Use the makefile for debug GI test * Update book * Tidy grpc and misc things * Apply discv5 fixes * Address other minor issues * Fix warnings * Attempt fix for addr parsing * Tidy validator config, CLIs * Tidy comments * Tidy signing, reduce ForkService duplication * Fail if skipping too many slots * Set default recent genesis time to 0 * Add custom http timeout to validator * Fix compile bug in node_test_rig * Remove old bootstrap flag from val CLI * Update docs * Tidy val client * Change val client log levels * Add comments, more validity checks * Fix compile error, add comments * Undo changes to eth2-libp2p/src * Reduce duplication of keypair generation * Add more logging for validator duties * Fix beacon_chain_sim, nitpicks * Fix compile error, minor nits * Update to use v0.9.2 version of deposit contract * Add efforts to automate eth1 testnet deployment * Fix lcli testnet deployer * Modify bn CLI to parse eth2_testnet_dir * Progress with account_manager deposit tools * Make account manager submit deposits * Add password option for submitting deposits * Allow custom deposit amount * Add long names to lcli clap * Add password option to lcli deploy command * Fix minor bugs whilst testing * Address Michael's comments * Add refund-deposit-contract to lcli * Use time instead of skip count for denying long skips * Improve logging for eth1 * Fix bug with validator services exiting on error * Drop the block cache after genesis * Modify eth1 testnet config * Improve eth1 logging * Make validator wait until genesis time * Fix bug in eth1 voting * Add more logging to eth1 voting * Handle errors in eth1 http module * Set SECONDS_PER_DAY to sensible minimum * Shorten delay before testnet start * Ensure eth1 block is produced without any votes * Improve eth1 logging * Fix broken tests in eth1 * Tidy code in rest_api * Fix failing test in deposit_contract * Make CLI args more consistent * Change validator/duties endpoint * Add time-based skip slot limiting * Add new error type missed in previous commit * Add log when waiting for genesis * Refactor beacon node CLI * Remove unused dep * Add lcli eth1-genesis command * Fix bug in master merge * Apply clippy lints to beacon node * Add support for YamlConfig in Eth2TestnetDir * Upgrade tesnet deposit contract version * Remove unnecessary logging and correct formatting * Add a hardcoded eth2 testnet config * Ensure http server flag works. Overwrite configs with flags. * Ensure boot nodes are loaded from testnet dir * Fix account manager CLI bugs * Fix bugs with beacon node cli * Allow testnet dir without boot nodes * Write genesis state as SSZ * Remove ---/n from the start of testnet_dir files * Set default libp2p address * Tidy account manager CLI, add logging * Add check to see if testnet dir exists * Apply reviewers suggestions * Add HeadTracker struct * Add fork choice persistence * Shorten slot time for simulator * Add the /beacon/heads API endpoint * Update hardcoded testnet * Add tests for BeaconChain persistence + fix bugs * Extend BeaconChain persistence testing * Ensure chain is finalized b4 persistence tests * Ensure boot_enr.yaml is include in binary * Refactor beacon_chain_sim * Move files about in beacon sim * Update beacon_chain_sim * Fix bug with deposit inclusion * Increase log in genesis service, fix todo * Tidy sim, fix broken rest_api tests * Fix more broken tests * Update testnet * Fix broken rest api test * Tidy account manager CLI * Use tempdir for account manager * Stop hardcoded testnet dir from creating dir * Rename Eth2TestnetDir to Eth2TestnetConfig * Change hardcoded -> hard_coded * Tidy account manager * Add log to account manager * Tidy, ensure head tracker is loaded from disk * Tidy beacon chain builder * Tidy eth1_chain * Adds log support for simulator * Revert "Adds log support for simulator" This reverts commit ec77c66a052350f551db145cf20f213823428dd3. * Adds log support for simulator * Tidy after self-review * Change default log level * Address Michael's delicious PR comments * Fix off-by-one in tests
2019-12-03 04:28:57 +00:00
}
/// This mutator will be run after the `initial_mutator`.
pub fn store_mutator(mut self, mutator: BoxedMutator<E, Hot, Cold>) -> Self {
assert!(self.store_mutator.is_none(), "store mutator already set");
self.store_mutator = Some(mutator);
self
}
Prepare for public testnet (#628) * Update to spec v0.9.0 * Update to v0.9.1 * Bump spec tags for v0.9.1 * Formatting, fix CI failures * Resolve accidental KeyPair merge conflict * Document new BeaconState functions * Add `validator` changes from `validator-to-rest` * Add initial (failing) REST api tests * Fix signature parsing * Add more tests * Refactor http router * Add working tests for publish beacon block * Add validator duties tests * Move account_manager under `lighthouse` binary * Unify logfile handling in `environment` crate. * Fix incorrect cache drops in `advance_caches` * Update fork choice for v0.9.1 * Add `deposit_contract` crate * Add progress on validator onboarding * Add unfinished attesation code * Update account manager CLI * Write eth1 data file as hex string * Integrate ValidatorDirectory with validator_client * Move ValidatorDirectory into validator_client * Clean up some FIXMEs * Add beacon_chain_sim * Fix a few docs/logs * Expand `beacon_chain_sim` * Fix spec for `beacon_chain_sim * More testing for api * Start work on attestation endpoint * Reject empty attestations * Allow attestations to genesis block * Add working tests for `rest_api` validator endpoint * Remove grpc from beacon_node * Start heavy refactor of validator client - Block production is working * Prune old validator client files * Start works on attestation service * Add attestation service to validator client * Use full pubkey for validator directories * Add validator duties post endpoint * Use par_iter for keypair generation * Use bulk duties request in validator client * Add version http endpoint tests * Add interop keys and startup wait * Ensure a prompt exit * Add duties pruning * Fix compile error in beacon node tests * Add github workflow * Modify rust.yaml * Modify gitlab actions * Add to CI file * Add sudo to CI npm install * Move cargo fmt to own job in tests * Fix cargo fmt in CI * Add rustup update before cargo fmt * Change name of CI job * Make other CI jobs require cargo fmt * Add CI badge * Remove gitlab and travis files * Add different http timeout for debug * Update docker file, use makefile in CI * Use make in the dockerfile, skip the test * Use the makefile for debug GI test * Update book * Tidy grpc and misc things * Apply discv5 fixes * Address other minor issues * Fix warnings * Attempt fix for addr parsing * Tidy validator config, CLIs * Tidy comments * Tidy signing, reduce ForkService duplication * Fail if skipping too many slots * Set default recent genesis time to 0 * Add custom http timeout to validator * Fix compile bug in node_test_rig * Remove old bootstrap flag from val CLI * Update docs * Tidy val client * Change val client log levels * Add comments, more validity checks * Fix compile error, add comments * Undo changes to eth2-libp2p/src * Reduce duplication of keypair generation * Add more logging for validator duties * Fix beacon_chain_sim, nitpicks * Fix compile error, minor nits * Update to use v0.9.2 version of deposit contract * Add efforts to automate eth1 testnet deployment * Fix lcli testnet deployer * Modify bn CLI to parse eth2_testnet_dir * Progress with account_manager deposit tools * Make account manager submit deposits * Add password option for submitting deposits * Allow custom deposit amount * Add long names to lcli clap * Add password option to lcli deploy command * Fix minor bugs whilst testing * Address Michael's comments * Add refund-deposit-contract to lcli * Use time instead of skip count for denying long skips * Improve logging for eth1 * Fix bug with validator services exiting on error * Drop the block cache after genesis * Modify eth1 testnet config * Improve eth1 logging * Make validator wait until genesis time * Fix bug in eth1 voting * Add more logging to eth1 voting * Handle errors in eth1 http module * Set SECONDS_PER_DAY to sensible minimum * Shorten delay before testnet start * Ensure eth1 block is produced without any votes * Improve eth1 logging * Fix broken tests in eth1 * Tidy code in rest_api * Fix failing test in deposit_contract * Make CLI args more consistent * Change validator/duties endpoint * Add time-based skip slot limiting * Add new error type missed in previous commit * Add log when waiting for genesis * Refactor beacon node CLI * Remove unused dep * Add lcli eth1-genesis command * Fix bug in master merge * Apply clippy lints to beacon node * Add support for YamlConfig in Eth2TestnetDir * Upgrade tesnet deposit contract version * Remove unnecessary logging and correct formatting * Add a hardcoded eth2 testnet config * Ensure http server flag works. Overwrite configs with flags. * Ensure boot nodes are loaded from testnet dir * Fix account manager CLI bugs * Fix bugs with beacon node cli * Allow testnet dir without boot nodes * Write genesis state as SSZ * Remove ---/n from the start of testnet_dir files * Set default libp2p address * Tidy account manager CLI, add logging * Add check to see if testnet dir exists * Apply reviewers suggestions * Add HeadTracker struct * Add fork choice persistence * Shorten slot time for simulator * Add the /beacon/heads API endpoint * Update hardcoded testnet * Add tests for BeaconChain persistence + fix bugs * Extend BeaconChain persistence testing * Ensure chain is finalized b4 persistence tests * Ensure boot_enr.yaml is include in binary * Refactor beacon_chain_sim * Move files about in beacon sim * Update beacon_chain_sim * Fix bug with deposit inclusion * Increase log in genesis service, fix todo * Tidy sim, fix broken rest_api tests * Fix more broken tests * Update testnet * Fix broken rest api test * Tidy account manager CLI * Use tempdir for account manager * Stop hardcoded testnet dir from creating dir * Rename Eth2TestnetDir to Eth2TestnetConfig * Change hardcoded -> hard_coded * Tidy account manager * Add log to account manager * Tidy, ensure head tracker is loaded from disk * Tidy beacon chain builder * Tidy eth1_chain * Adds log support for simulator * Revert "Adds log support for simulator" This reverts commit ec77c66a052350f551db145cf20f213823428dd3. * Adds log support for simulator * Tidy after self-review * Change default log level * Address Michael's delicious PR comments * Fix off-by-one in tests
2019-12-03 04:28:57 +00:00
/// Purposefully replace the `store_mutator`.
pub fn override_store_mutator(mut self, mutator: BoxedMutator<E, Hot, Cold>) -> Self {
assert!(self.store_mutator.is_some(), "store mutator not set");
self.store_mutator = Some(mutator);
self
}
pub fn chain_config(mut self, chain_config: ChainConfig) -> Self {
self.chain_config = Some(chain_config);
self
}
pub fn execution_layer(mut self, urls: &[&str]) -> Self {
assert!(
self.execution_layer.is_none(),
"execution layer already defined"
);
let urls: Vec<SensitiveUrl> = urls
.iter()
.map(|s| SensitiveUrl::parse(*s))
.collect::<Result<_, _>>()
.unwrap();
let config = execution_layer::Config {
execution_endpoints: urls,
secret_files: vec![],
suggested_fee_recipient: Some(Address::repeat_byte(42)),
..Default::default()
};
let execution_layer = ExecutionLayer::from_config(
config,
self.runtime.task_executor.clone(),
self.log.clone(),
)
.unwrap();
self.execution_layer = Some(execution_layer);
self
}
pub fn mock_execution_layer(mut self) -> Self {
let spec = self.spec.clone().expect("cannot build without spec");
let mock = MockExecutionLayer::new(
self.runtime.task_executor.clone(),
spec.terminal_total_difficulty,
DEFAULT_TERMINAL_BLOCK,
spec.terminal_block_hash,
spec.terminal_block_hash_activation_epoch,
Some(JwtKey::from_slice(&DEFAULT_JWT_SECRET).unwrap()),
None,
);
self.execution_layer = Some(mock.el.clone());
self.mock_execution_layer = Some(mock);
self
}
/// Instruct the mock execution engine to always return a "valid" response to any payload it is
/// asked to execute.
pub fn mock_execution_layer_all_payloads_valid(self) -> Self {
self.mock_execution_layer
.as_ref()
.expect("requires mock execution layer")
.server
.all_payloads_valid();
self
}
pub fn build(self) -> BeaconChainHarness<BaseHarnessType<E, Hot, Cold>> {
let (shutdown_tx, shutdown_receiver) = futures::channel::mpsc::channel(1);
Prepare for public testnet (#628) * Update to spec v0.9.0 * Update to v0.9.1 * Bump spec tags for v0.9.1 * Formatting, fix CI failures * Resolve accidental KeyPair merge conflict * Document new BeaconState functions * Add `validator` changes from `validator-to-rest` * Add initial (failing) REST api tests * Fix signature parsing * Add more tests * Refactor http router * Add working tests for publish beacon block * Add validator duties tests * Move account_manager under `lighthouse` binary * Unify logfile handling in `environment` crate. * Fix incorrect cache drops in `advance_caches` * Update fork choice for v0.9.1 * Add `deposit_contract` crate * Add progress on validator onboarding * Add unfinished attesation code * Update account manager CLI * Write eth1 data file as hex string * Integrate ValidatorDirectory with validator_client * Move ValidatorDirectory into validator_client * Clean up some FIXMEs * Add beacon_chain_sim * Fix a few docs/logs * Expand `beacon_chain_sim` * Fix spec for `beacon_chain_sim * More testing for api * Start work on attestation endpoint * Reject empty attestations * Allow attestations to genesis block * Add working tests for `rest_api` validator endpoint * Remove grpc from beacon_node * Start heavy refactor of validator client - Block production is working * Prune old validator client files * Start works on attestation service * Add attestation service to validator client * Use full pubkey for validator directories * Add validator duties post endpoint * Use par_iter for keypair generation * Use bulk duties request in validator client * Add version http endpoint tests * Add interop keys and startup wait * Ensure a prompt exit * Add duties pruning * Fix compile error in beacon node tests * Add github workflow * Modify rust.yaml * Modify gitlab actions * Add to CI file * Add sudo to CI npm install * Move cargo fmt to own job in tests * Fix cargo fmt in CI * Add rustup update before cargo fmt * Change name of CI job * Make other CI jobs require cargo fmt * Add CI badge * Remove gitlab and travis files * Add different http timeout for debug * Update docker file, use makefile in CI * Use make in the dockerfile, skip the test * Use the makefile for debug GI test * Update book * Tidy grpc and misc things * Apply discv5 fixes * Address other minor issues * Fix warnings * Attempt fix for addr parsing * Tidy validator config, CLIs * Tidy comments * Tidy signing, reduce ForkService duplication * Fail if skipping too many slots * Set default recent genesis time to 0 * Add custom http timeout to validator * Fix compile bug in node_test_rig * Remove old bootstrap flag from val CLI * Update docs * Tidy val client * Change val client log levels * Add comments, more validity checks * Fix compile error, add comments * Undo changes to eth2-libp2p/src * Reduce duplication of keypair generation * Add more logging for validator duties * Fix beacon_chain_sim, nitpicks * Fix compile error, minor nits * Update to use v0.9.2 version of deposit contract * Add efforts to automate eth1 testnet deployment * Fix lcli testnet deployer * Modify bn CLI to parse eth2_testnet_dir * Progress with account_manager deposit tools * Make account manager submit deposits * Add password option for submitting deposits * Allow custom deposit amount * Add long names to lcli clap * Add password option to lcli deploy command * Fix minor bugs whilst testing * Address Michael's comments * Add refund-deposit-contract to lcli * Use time instead of skip count for denying long skips * Improve logging for eth1 * Fix bug with validator services exiting on error * Drop the block cache after genesis * Modify eth1 testnet config * Improve eth1 logging * Make validator wait until genesis time * Fix bug in eth1 voting * Add more logging to eth1 voting * Handle errors in eth1 http module * Set SECONDS_PER_DAY to sensible minimum * Shorten delay before testnet start * Ensure eth1 block is produced without any votes * Improve eth1 logging * Fix broken tests in eth1 * Tidy code in rest_api * Fix failing test in deposit_contract * Make CLI args more consistent * Change validator/duties endpoint * Add time-based skip slot limiting * Add new error type missed in previous commit * Add log when waiting for genesis * Refactor beacon node CLI * Remove unused dep * Add lcli eth1-genesis command * Fix bug in master merge * Apply clippy lints to beacon node * Add support for YamlConfig in Eth2TestnetDir * Upgrade tesnet deposit contract version * Remove unnecessary logging and correct formatting * Add a hardcoded eth2 testnet config * Ensure http server flag works. Overwrite configs with flags. * Ensure boot nodes are loaded from testnet dir * Fix account manager CLI bugs * Fix bugs with beacon node cli * Allow testnet dir without boot nodes * Write genesis state as SSZ * Remove ---/n from the start of testnet_dir files * Set default libp2p address * Tidy account manager CLI, add logging * Add check to see if testnet dir exists * Apply reviewers suggestions * Add HeadTracker struct * Add fork choice persistence * Shorten slot time for simulator * Add the /beacon/heads API endpoint * Update hardcoded testnet * Add tests for BeaconChain persistence + fix bugs * Extend BeaconChain persistence testing * Ensure chain is finalized b4 persistence tests * Ensure boot_enr.yaml is include in binary * Refactor beacon_chain_sim * Move files about in beacon sim * Update beacon_chain_sim * Fix bug with deposit inclusion * Increase log in genesis service, fix todo * Tidy sim, fix broken rest_api tests * Fix more broken tests * Update testnet * Fix broken rest api test * Tidy account manager CLI * Use tempdir for account manager * Stop hardcoded testnet dir from creating dir * Rename Eth2TestnetDir to Eth2TestnetConfig * Change hardcoded -> hard_coded * Tidy account manager * Add log to account manager * Tidy, ensure head tracker is loaded from disk * Tidy beacon chain builder * Tidy eth1_chain * Adds log support for simulator * Revert "Adds log support for simulator" This reverts commit ec77c66a052350f551db145cf20f213823428dd3. * Adds log support for simulator * Tidy after self-review * Change default log level * Address Michael's delicious PR comments * Fix off-by-one in tests
2019-12-03 04:28:57 +00:00
let log = self.log;
let spec = self.spec.expect("cannot build without spec");
let seconds_per_slot = spec.seconds_per_slot;
let validator_keypairs = self
.validator_keypairs
.expect("cannot build without validator keypairs");
let mut builder = BeaconChainBuilder::new(self.eth_spec_instance)
Prepare for public testnet (#628) * Update to spec v0.9.0 * Update to v0.9.1 * Bump spec tags for v0.9.1 * Formatting, fix CI failures * Resolve accidental KeyPair merge conflict * Document new BeaconState functions * Add `validator` changes from `validator-to-rest` * Add initial (failing) REST api tests * Fix signature parsing * Add more tests * Refactor http router * Add working tests for publish beacon block * Add validator duties tests * Move account_manager under `lighthouse` binary * Unify logfile handling in `environment` crate. * Fix incorrect cache drops in `advance_caches` * Update fork choice for v0.9.1 * Add `deposit_contract` crate * Add progress on validator onboarding * Add unfinished attesation code * Update account manager CLI * Write eth1 data file as hex string * Integrate ValidatorDirectory with validator_client * Move ValidatorDirectory into validator_client * Clean up some FIXMEs * Add beacon_chain_sim * Fix a few docs/logs * Expand `beacon_chain_sim` * Fix spec for `beacon_chain_sim * More testing for api * Start work on attestation endpoint * Reject empty attestations * Allow attestations to genesis block * Add working tests for `rest_api` validator endpoint * Remove grpc from beacon_node * Start heavy refactor of validator client - Block production is working * Prune old validator client files * Start works on attestation service * Add attestation service to validator client * Use full pubkey for validator directories * Add validator duties post endpoint * Use par_iter for keypair generation * Use bulk duties request in validator client * Add version http endpoint tests * Add interop keys and startup wait * Ensure a prompt exit * Add duties pruning * Fix compile error in beacon node tests * Add github workflow * Modify rust.yaml * Modify gitlab actions * Add to CI file * Add sudo to CI npm install * Move cargo fmt to own job in tests * Fix cargo fmt in CI * Add rustup update before cargo fmt * Change name of CI job * Make other CI jobs require cargo fmt * Add CI badge * Remove gitlab and travis files * Add different http timeout for debug * Update docker file, use makefile in CI * Use make in the dockerfile, skip the test * Use the makefile for debug GI test * Update book * Tidy grpc and misc things * Apply discv5 fixes * Address other minor issues * Fix warnings * Attempt fix for addr parsing * Tidy validator config, CLIs * Tidy comments * Tidy signing, reduce ForkService duplication * Fail if skipping too many slots * Set default recent genesis time to 0 * Add custom http timeout to validator * Fix compile bug in node_test_rig * Remove old bootstrap flag from val CLI * Update docs * Tidy val client * Change val client log levels * Add comments, more validity checks * Fix compile error, add comments * Undo changes to eth2-libp2p/src * Reduce duplication of keypair generation * Add more logging for validator duties * Fix beacon_chain_sim, nitpicks * Fix compile error, minor nits * Update to use v0.9.2 version of deposit contract * Add efforts to automate eth1 testnet deployment * Fix lcli testnet deployer * Modify bn CLI to parse eth2_testnet_dir * Progress with account_manager deposit tools * Make account manager submit deposits * Add password option for submitting deposits * Allow custom deposit amount * Add long names to lcli clap * Add password option to lcli deploy command * Fix minor bugs whilst testing * Address Michael's comments * Add refund-deposit-contract to lcli * Use time instead of skip count for denying long skips * Improve logging for eth1 * Fix bug with validator services exiting on error * Drop the block cache after genesis * Modify eth1 testnet config * Improve eth1 logging * Make validator wait until genesis time * Fix bug in eth1 voting * Add more logging to eth1 voting * Handle errors in eth1 http module * Set SECONDS_PER_DAY to sensible minimum * Shorten delay before testnet start * Ensure eth1 block is produced without any votes * Improve eth1 logging * Fix broken tests in eth1 * Tidy code in rest_api * Fix failing test in deposit_contract * Make CLI args more consistent * Change validator/duties endpoint * Add time-based skip slot limiting * Add new error type missed in previous commit * Add log when waiting for genesis * Refactor beacon node CLI * Remove unused dep * Add lcli eth1-genesis command * Fix bug in master merge * Apply clippy lints to beacon node * Add support for YamlConfig in Eth2TestnetDir * Upgrade tesnet deposit contract version * Remove unnecessary logging and correct formatting * Add a hardcoded eth2 testnet config * Ensure http server flag works. Overwrite configs with flags. * Ensure boot nodes are loaded from testnet dir * Fix account manager CLI bugs * Fix bugs with beacon node cli * Allow testnet dir without boot nodes * Write genesis state as SSZ * Remove ---/n from the start of testnet_dir files * Set default libp2p address * Tidy account manager CLI, add logging * Add check to see if testnet dir exists * Apply reviewers suggestions * Add HeadTracker struct * Add fork choice persistence * Shorten slot time for simulator * Add the /beacon/heads API endpoint * Update hardcoded testnet * Add tests for BeaconChain persistence + fix bugs * Extend BeaconChain persistence testing * Ensure chain is finalized b4 persistence tests * Ensure boot_enr.yaml is include in binary * Refactor beacon_chain_sim * Move files about in beacon sim * Update beacon_chain_sim * Fix bug with deposit inclusion * Increase log in genesis service, fix todo * Tidy sim, fix broken rest_api tests * Fix more broken tests * Update testnet * Fix broken rest api test * Tidy account manager CLI * Use tempdir for account manager * Stop hardcoded testnet dir from creating dir * Rename Eth2TestnetDir to Eth2TestnetConfig * Change hardcoded -> hard_coded * Tidy account manager * Add log to account manager * Tidy, ensure head tracker is loaded from disk * Tidy beacon chain builder * Tidy eth1_chain * Adds log support for simulator * Revert "Adds log support for simulator" This reverts commit ec77c66a052350f551db145cf20f213823428dd3. * Adds log support for simulator * Tidy after self-review * Change default log level * Address Michael's delicious PR comments * Fix off-by-one in tests
2019-12-03 04:28:57 +00:00
.logger(log.clone())
.custom_spec(spec)
.store(self.store.expect("cannot build without store"))
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
.store_migrator_config(MigratorConfig::default().blocking())
.task_executor(self.runtime.task_executor.clone())
.execution_layer(self.execution_layer)
Prepare for public testnet (#628) * Update to spec v0.9.0 * Update to v0.9.1 * Bump spec tags for v0.9.1 * Formatting, fix CI failures * Resolve accidental KeyPair merge conflict * Document new BeaconState functions * Add `validator` changes from `validator-to-rest` * Add initial (failing) REST api tests * Fix signature parsing * Add more tests * Refactor http router * Add working tests for publish beacon block * Add validator duties tests * Move account_manager under `lighthouse` binary * Unify logfile handling in `environment` crate. * Fix incorrect cache drops in `advance_caches` * Update fork choice for v0.9.1 * Add `deposit_contract` crate * Add progress on validator onboarding * Add unfinished attesation code * Update account manager CLI * Write eth1 data file as hex string * Integrate ValidatorDirectory with validator_client * Move ValidatorDirectory into validator_client * Clean up some FIXMEs * Add beacon_chain_sim * Fix a few docs/logs * Expand `beacon_chain_sim` * Fix spec for `beacon_chain_sim * More testing for api * Start work on attestation endpoint * Reject empty attestations * Allow attestations to genesis block * Add working tests for `rest_api` validator endpoint * Remove grpc from beacon_node * Start heavy refactor of validator client - Block production is working * Prune old validator client files * Start works on attestation service * Add attestation service to validator client * Use full pubkey for validator directories * Add validator duties post endpoint * Use par_iter for keypair generation * Use bulk duties request in validator client * Add version http endpoint tests * Add interop keys and startup wait * Ensure a prompt exit * Add duties pruning * Fix compile error in beacon node tests * Add github workflow * Modify rust.yaml * Modify gitlab actions * Add to CI file * Add sudo to CI npm install * Move cargo fmt to own job in tests * Fix cargo fmt in CI * Add rustup update before cargo fmt * Change name of CI job * Make other CI jobs require cargo fmt * Add CI badge * Remove gitlab and travis files * Add different http timeout for debug * Update docker file, use makefile in CI * Use make in the dockerfile, skip the test * Use the makefile for debug GI test * Update book * Tidy grpc and misc things * Apply discv5 fixes * Address other minor issues * Fix warnings * Attempt fix for addr parsing * Tidy validator config, CLIs * Tidy comments * Tidy signing, reduce ForkService duplication * Fail if skipping too many slots * Set default recent genesis time to 0 * Add custom http timeout to validator * Fix compile bug in node_test_rig * Remove old bootstrap flag from val CLI * Update docs * Tidy val client * Change val client log levels * Add comments, more validity checks * Fix compile error, add comments * Undo changes to eth2-libp2p/src * Reduce duplication of keypair generation * Add more logging for validator duties * Fix beacon_chain_sim, nitpicks * Fix compile error, minor nits * Update to use v0.9.2 version of deposit contract * Add efforts to automate eth1 testnet deployment * Fix lcli testnet deployer * Modify bn CLI to parse eth2_testnet_dir * Progress with account_manager deposit tools * Make account manager submit deposits * Add password option for submitting deposits * Allow custom deposit amount * Add long names to lcli clap * Add password option to lcli deploy command * Fix minor bugs whilst testing * Address Michael's comments * Add refund-deposit-contract to lcli * Use time instead of skip count for denying long skips * Improve logging for eth1 * Fix bug with validator services exiting on error * Drop the block cache after genesis * Modify eth1 testnet config * Improve eth1 logging * Make validator wait until genesis time * Fix bug in eth1 voting * Add more logging to eth1 voting * Handle errors in eth1 http module * Set SECONDS_PER_DAY to sensible minimum * Shorten delay before testnet start * Ensure eth1 block is produced without any votes * Improve eth1 logging * Fix broken tests in eth1 * Tidy code in rest_api * Fix failing test in deposit_contract * Make CLI args more consistent * Change validator/duties endpoint * Add time-based skip slot limiting * Add new error type missed in previous commit * Add log when waiting for genesis * Refactor beacon node CLI * Remove unused dep * Add lcli eth1-genesis command * Fix bug in master merge * Apply clippy lints to beacon node * Add support for YamlConfig in Eth2TestnetDir * Upgrade tesnet deposit contract version * Remove unnecessary logging and correct formatting * Add a hardcoded eth2 testnet config * Ensure http server flag works. Overwrite configs with flags. * Ensure boot nodes are loaded from testnet dir * Fix account manager CLI bugs * Fix bugs with beacon node cli * Allow testnet dir without boot nodes * Write genesis state as SSZ * Remove ---/n from the start of testnet_dir files * Set default libp2p address * Tidy account manager CLI, add logging * Add check to see if testnet dir exists * Apply reviewers suggestions * Add HeadTracker struct * Add fork choice persistence * Shorten slot time for simulator * Add the /beacon/heads API endpoint * Update hardcoded testnet * Add tests for BeaconChain persistence + fix bugs * Extend BeaconChain persistence testing * Ensure chain is finalized b4 persistence tests * Ensure boot_enr.yaml is include in binary * Refactor beacon_chain_sim * Move files about in beacon sim * Update beacon_chain_sim * Fix bug with deposit inclusion * Increase log in genesis service, fix todo * Tidy sim, fix broken rest_api tests * Fix more broken tests * Update testnet * Fix broken rest api test * Tidy account manager CLI * Use tempdir for account manager * Stop hardcoded testnet dir from creating dir * Rename Eth2TestnetDir to Eth2TestnetConfig * Change hardcoded -> hard_coded * Tidy account manager * Add log to account manager * Tidy, ensure head tracker is loaded from disk * Tidy beacon chain builder * Tidy eth1_chain * Adds log support for simulator * Revert "Adds log support for simulator" This reverts commit ec77c66a052350f551db145cf20f213823428dd3. * Adds log support for simulator * Tidy after self-review * Change default log level * Address Michael's delicious PR comments * Fix off-by-one in tests
2019-12-03 04:28:57 +00:00
.dummy_eth1_backend()
.expect("should build dummy backend")
.shutdown_sender(shutdown_tx)
.chain_config(self.chain_config.unwrap_or_default())
.event_handler(Some(ServerSentEventHandler::new_with_capacity(
log.clone(),
5,
)))
.monitor_validators(true, vec![], log);
builder = if let Some(mutator) = self.initial_mutator {
mutator(builder)
} else {
builder
};
builder = if let Some(mutator) = self.store_mutator {
mutator(builder)
} else {
builder
};
// Initialize the slot clock only if it hasn't already been initialized.
builder = if builder.get_slot_clock().is_none() {
builder
.testing_slot_clock(Duration::from_secs(seconds_per_slot))
.expect("should configure testing slot clock")
} else {
builder
};
let chain = builder.build().expect("should build");
BeaconChainHarness {
spec: chain.spec.clone(),
chain: Arc::new(chain),
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
validator_keypairs,
shutdown_receiver: Arc::new(Mutex::new(shutdown_receiver)),
runtime: self.runtime,
mock_execution_layer: self.mock_execution_layer,
rng: make_rng(),
}
}
}
/// A testing harness which can instantiate a `BeaconChain` and populate it with blocks and
/// attestations.
///
/// Used for testing.
pub struct BeaconChainHarness<T: BeaconChainTypes> {
pub validator_keypairs: Vec<Keypair>,
pub chain: Arc<BeaconChain<T>>,
pub spec: ChainSpec,
pub shutdown_receiver: Arc<Mutex<Receiver<ShutdownReason>>>,
pub runtime: TestRuntime,
pub mock_execution_layer: Option<MockExecutionLayer<T::EthSpec>>,
pub rng: Mutex<StdRng>,
}
pub type HarnessAttestations<E> = Vec<(
Vec<(Attestation<E>, SubnetId)>,
Option<SignedAggregateAndProof<E>>,
)>;
pub type HarnessSyncContributions<E> = Vec<(
Vec<(SyncCommitteeMessage, usize)>,
Option<SignedContributionAndProof<E>>,
)>;
impl<E, Hot, Cold> BeaconChainHarness<BaseHarnessType<E, Hot, Cold>>
where
E: EthSpec,
Hot: ItemStore<E>,
Cold: ItemStore<E>,
{
pub fn builder(eth_spec_instance: E) -> Builder<BaseHarnessType<E, Hot, Cold>> {
Builder::new(eth_spec_instance)
}
2019-06-16 06:24:33 +00:00
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
pub fn logger(&self) -> &slog::Logger {
&self.chain.log
}
pub fn execution_block_generator(&self) -> RwLockWriteGuard<'_, ExecutionBlockGenerator<E>> {
self.mock_execution_layer
.as_ref()
.expect("harness was not built with mock execution layer")
.server
.execution_block_generator()
}
pub fn get_all_validators(&self) -> Vec<usize> {
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
(0..self.validator_keypairs.len()).collect()
v0.12 fork choice update (#1229) * Incomplete scraps * Add progress on new fork choice impl * Further progress * First complete compiling version * Remove chain reference * Add new lmd_ghost crate * Start integrating into beacon chain * Update `milagro_bls` to new release (#1183) * Update milagro_bls to new release Signed-off-by: Kirk Baird <baird.k@outlook.com> * Tidy up fake cryptos Signed-off-by: Kirk Baird <baird.k@outlook.com> * move SecretHash to bls and put plaintext back Signed-off-by: Kirk Baird <baird.k@outlook.com> * Update state processing for v0.12 * Fix EF test runners for v0.12 * Fix some tests * Fix broken attestation verification test * More test fixes * Rough beacon chain impl working * Remove fork_choice_2 * Remove checkpoint manager * Half finished ssz impl * Add missed file * Add persistence * Tidy, fix some compile errors * Remove RwLock from ProtoArrayForkChoice * Fix store-based compile errors * Add comments, tidy * Move function out of ForkChoice struct * Start testing * More testing * Fix compile error * Tidy beacon_chain::fork_choice * Queue attestations from the current slot * Allow fork choice to handle prior-to-genesis start * Improve error granularity * Test attestation dequeuing * Process attestations during block * Store target root in fork choice * Move fork choice verification into new crate * Update tests * Consensus updates for v0.12 (#1228) * Update state processing for v0.12 * Fix EF test runners for v0.12 * Fix some tests * Fix broken attestation verification test * More test fixes * Fix typo found in review * Add `Block` struct to ProtoArray * Start fixing get_ancestor * Add rough progress on testing * Get fork choice tests working * Progress with testing * Fix partialeq impl * Move slot clock from fc_store * Improve testing * Add testing for best justified * Add clone back to SystemTimeSlotClock * Add balances test * Start adding balances cache again * Wire-in balances cache * Improve tests * Remove commented-out tests * Remove beacon_chain::ForkChoice * Rename crates * Update wider codebase to new fork_choice layout * Move advance_slot in test harness * Tidy ForkChoice::update_time * Fix verification tests * Fix compile error with iter::once * Fix fork choice tests * Ensure block attestations are processed * Fix failing beacon_chain tests * Add first invalid block check * Add finalized block check * Progress with testing, new store builder * Add fixes to get_ancestor * Fix old genesis justification test * Fix remaining fork choice tests * Change root iteration method * Move on_verified_block * Remove unused method * Start adding attestation verification tests * Add invalid ffg target test * Add target epoch test * Add queued attestation test * Remove old fork choice verification tests * Tidy, add test * Move fork choice lock drop * Rename BeaconForkChoiceStore * Add comments, tidy BeaconForkChoiceStore * Update metrics, rename fork_choice_store.rs * Remove genesis_block_root from ForkChoice * Tidy * Update fork_choice comments * Tidy, add comments * Tidy, simplify ForkChoice, fix compile issue * Tidy, removed dead file * Increase http request timeout * Fix failing rest_api test * Set HTTP timeout back to 5s * Apply fix to get_ancestor * Address Michael's comments * Fix typo * Revert "Fix broken attestation verification test" This reverts commit 722cdc903b12611de27916a57eeecfa3224f2279. Co-authored-by: Kirk Baird <baird.k@outlook.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-06-17 01:10:22 +00:00
}
pub fn slots_per_epoch(&self) -> u64 {
E::slots_per_epoch()
}
pub fn epoch_start_slot(&self, epoch: u64) -> u64 {
let epoch = Epoch::new(epoch);
epoch.start_slot(E::slots_per_epoch()).into()
v0.12 fork choice update (#1229) * Incomplete scraps * Add progress on new fork choice impl * Further progress * First complete compiling version * Remove chain reference * Add new lmd_ghost crate * Start integrating into beacon chain * Update `milagro_bls` to new release (#1183) * Update milagro_bls to new release Signed-off-by: Kirk Baird <baird.k@outlook.com> * Tidy up fake cryptos Signed-off-by: Kirk Baird <baird.k@outlook.com> * move SecretHash to bls and put plaintext back Signed-off-by: Kirk Baird <baird.k@outlook.com> * Update state processing for v0.12 * Fix EF test runners for v0.12 * Fix some tests * Fix broken attestation verification test * More test fixes * Rough beacon chain impl working * Remove fork_choice_2 * Remove checkpoint manager * Half finished ssz impl * Add missed file * Add persistence * Tidy, fix some compile errors * Remove RwLock from ProtoArrayForkChoice * Fix store-based compile errors * Add comments, tidy * Move function out of ForkChoice struct * Start testing * More testing * Fix compile error * Tidy beacon_chain::fork_choice * Queue attestations from the current slot * Allow fork choice to handle prior-to-genesis start * Improve error granularity * Test attestation dequeuing * Process attestations during block * Store target root in fork choice * Move fork choice verification into new crate * Update tests * Consensus updates for v0.12 (#1228) * Update state processing for v0.12 * Fix EF test runners for v0.12 * Fix some tests * Fix broken attestation verification test * More test fixes * Fix typo found in review * Add `Block` struct to ProtoArray * Start fixing get_ancestor * Add rough progress on testing * Get fork choice tests working * Progress with testing * Fix partialeq impl * Move slot clock from fc_store * Improve testing * Add testing for best justified * Add clone back to SystemTimeSlotClock * Add balances test * Start adding balances cache again * Wire-in balances cache * Improve tests * Remove commented-out tests * Remove beacon_chain::ForkChoice * Rename crates * Update wider codebase to new fork_choice layout * Move advance_slot in test harness * Tidy ForkChoice::update_time * Fix verification tests * Fix compile error with iter::once * Fix fork choice tests * Ensure block attestations are processed * Fix failing beacon_chain tests * Add first invalid block check * Add finalized block check * Progress with testing, new store builder * Add fixes to get_ancestor * Fix old genesis justification test * Fix remaining fork choice tests * Change root iteration method * Move on_verified_block * Remove unused method * Start adding attestation verification tests * Add invalid ffg target test * Add target epoch test * Add queued attestation test * Remove old fork choice verification tests * Tidy, add test * Move fork choice lock drop * Rename BeaconForkChoiceStore * Add comments, tidy BeaconForkChoiceStore * Update metrics, rename fork_choice_store.rs * Remove genesis_block_root from ForkChoice * Tidy * Update fork_choice comments * Tidy, add comments * Tidy, simplify ForkChoice, fix compile issue * Tidy, removed dead file * Increase http request timeout * Fix failing rest_api test * Set HTTP timeout back to 5s * Apply fix to get_ancestor * Address Michael's comments * Fix typo * Revert "Fix broken attestation verification test" This reverts commit 722cdc903b12611de27916a57eeecfa3224f2279. Co-authored-by: Kirk Baird <baird.k@outlook.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-06-17 01:10:22 +00:00
}
pub fn shutdown_reasons(&self) -> Vec<ShutdownReason> {
let mutex = self.shutdown_receiver.clone();
let mut receiver = mutex.lock();
std::iter::from_fn(move || match receiver.try_next() {
Ok(Some(s)) => Some(s),
Ok(None) => panic!("shutdown sender dropped"),
Err(_) => None,
})
.collect()
}
pub fn get_current_state(&self) -> BeaconState<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
self.chain.head_beacon_state_cloned()
v0.12 fork choice update (#1229) * Incomplete scraps * Add progress on new fork choice impl * Further progress * First complete compiling version * Remove chain reference * Add new lmd_ghost crate * Start integrating into beacon chain * Update `milagro_bls` to new release (#1183) * Update milagro_bls to new release Signed-off-by: Kirk Baird <baird.k@outlook.com> * Tidy up fake cryptos Signed-off-by: Kirk Baird <baird.k@outlook.com> * move SecretHash to bls and put plaintext back Signed-off-by: Kirk Baird <baird.k@outlook.com> * Update state processing for v0.12 * Fix EF test runners for v0.12 * Fix some tests * Fix broken attestation verification test * More test fixes * Rough beacon chain impl working * Remove fork_choice_2 * Remove checkpoint manager * Half finished ssz impl * Add missed file * Add persistence * Tidy, fix some compile errors * Remove RwLock from ProtoArrayForkChoice * Fix store-based compile errors * Add comments, tidy * Move function out of ForkChoice struct * Start testing * More testing * Fix compile error * Tidy beacon_chain::fork_choice * Queue attestations from the current slot * Allow fork choice to handle prior-to-genesis start * Improve error granularity * Test attestation dequeuing * Process attestations during block * Store target root in fork choice * Move fork choice verification into new crate * Update tests * Consensus updates for v0.12 (#1228) * Update state processing for v0.12 * Fix EF test runners for v0.12 * Fix some tests * Fix broken attestation verification test * More test fixes * Fix typo found in review * Add `Block` struct to ProtoArray * Start fixing get_ancestor * Add rough progress on testing * Get fork choice tests working * Progress with testing * Fix partialeq impl * Move slot clock from fc_store * Improve testing * Add testing for best justified * Add clone back to SystemTimeSlotClock * Add balances test * Start adding balances cache again * Wire-in balances cache * Improve tests * Remove commented-out tests * Remove beacon_chain::ForkChoice * Rename crates * Update wider codebase to new fork_choice layout * Move advance_slot in test harness * Tidy ForkChoice::update_time * Fix verification tests * Fix compile error with iter::once * Fix fork choice tests * Ensure block attestations are processed * Fix failing beacon_chain tests * Add first invalid block check * Add finalized block check * Progress with testing, new store builder * Add fixes to get_ancestor * Fix old genesis justification test * Fix remaining fork choice tests * Change root iteration method * Move on_verified_block * Remove unused method * Start adding attestation verification tests * Add invalid ffg target test * Add target epoch test * Add queued attestation test * Remove old fork choice verification tests * Tidy, add test * Move fork choice lock drop * Rename BeaconForkChoiceStore * Add comments, tidy BeaconForkChoiceStore * Update metrics, rename fork_choice_store.rs * Remove genesis_block_root from ForkChoice * Tidy * Update fork_choice comments * Tidy, add comments * Tidy, simplify ForkChoice, fix compile issue * Tidy, removed dead file * Increase http request timeout * Fix failing rest_api test * Set HTTP timeout back to 5s * Apply fix to get_ancestor * Address Michael's comments * Fix typo * Revert "Fix broken attestation verification test" This reverts commit 722cdc903b12611de27916a57eeecfa3224f2279. Co-authored-by: Kirk Baird <baird.k@outlook.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-06-17 01:10:22 +00:00
}
pub fn get_timestamp_at_slot(&self) -> u64 {
let state = self.get_current_state();
compute_timestamp_at_slot(&state, &self.spec).unwrap()
}
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
pub fn get_current_state_and_root(&self) -> (BeaconState<E>, Hash256) {
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 = self.chain.head_snapshot();
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 = head.beacon_state_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
(
head.beacon_state.clone_with_only_committee_caches(),
state_root,
)
}
pub fn head_slot(&self) -> Slot {
self.chain.canonical_head.cached_head().head_slot()
}
pub fn head_block_root(&self) -> Hash256 {
self.chain.canonical_head.cached_head().head_block_root()
}
pub fn finalized_checkpoint(&self) -> Checkpoint {
self.chain
.canonical_head
.cached_head()
.finalized_checkpoint()
}
pub fn justified_checkpoint(&self) -> Checkpoint {
self.chain
.canonical_head
.cached_head()
.justified_checkpoint()
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
}
pub fn get_current_slot(&self) -> 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
self.chain.slot().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
pub fn get_block(
&self,
block_hash: SignedBeaconBlockHash,
) -> Option<SignedBeaconBlock<E, BlindedPayload<E>>> {
self.chain.get_blinded_block(&block_hash.into()).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
}
pub fn block_exists(&self, block_hash: SignedBeaconBlockHash) -> bool {
self.get_block(block_hash).is_some()
}
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
pub fn get_hot_state(&self, state_hash: BeaconStateHash) -> Option<BeaconState<E>> {
self.chain
.store
.load_hot_state(&state_hash.into(), StateRootStrategy::Accurate)
.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
pub fn get_cold_state(&self, state_hash: BeaconStateHash) -> Option<BeaconState<E>> {
self.chain
.store
.load_cold_state(&state_hash.into())
.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
}
pub fn hot_state_exists(&self, state_hash: BeaconStateHash) -> bool {
self.get_hot_state(state_hash).is_some()
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
}
pub fn cold_state_exists(&self, state_hash: BeaconStateHash) -> bool {
self.get_cold_state(state_hash).is_some()
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
}
pub fn is_skipped_slot(&self, state: &BeaconState<E>, slot: Slot) -> bool {
state.get_block_root(slot).unwrap() == state.get_block_root(slot - 1).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
}
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
pub async fn make_block(
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
&self,
2019-06-20 00:59:19 +00:00
mut state: BeaconState<E>,
slot: Slot,
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
) -> (SignedBeaconBlock<E>, BeaconState<E>) {
assert_ne!(slot, 0, "can't produce a block at slot 0");
assert!(slot >= state.slot());
2019-06-20 00:59:19 +00:00
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
complete_state_advance(&mut state, None, slot, &self.spec)
.expect("should be able to advance state to slot");
2019-06-20 00:59:19 +00:00
state
.build_all_caches(&self.spec)
.expect("should build caches");
let proposer_index = state.get_beacon_proposer_index(slot, &self.spec).unwrap();
2019-06-16 06:24:33 +00:00
// If we produce two blocks for the same slot, they hash up to the same value and
// BeaconChain errors out with `BlockIsAlreadyKnown`. Vary the graffiti so that we produce
// different blocks each time.
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 graffiti = Graffiti::from(self.rng.lock().gen::<[u8; 32]>());
2019-06-16 06:24:33 +00:00
Run fork choice before block proposal (#3168) ## Issue Addressed Upcoming spec change https://github.com/ethereum/consensus-specs/pull/2878 ## Proposed Changes 1. Run fork choice at the start of every slot, and wait for this run to complete before proposing a block. 2. As an optimisation, also run fork choice 3/4 of the way through the slot (at 9s), _dequeueing attestations for the next slot_. 3. Remove the fork choice run from the state advance timer that occurred before advancing the state. ## Additional Info ### Block Proposal Accuracy This change makes us more likely to propose on top of the correct head in the presence of re-orgs with proposer boost in play. The main scenario that this change is designed to address is described in the linked spec issue. ### Attestation Accuracy This change _also_ makes us more likely to attest to the correct head. Currently in the case of a skipped slot at `slot` we only run fork choice 9s into `slot - 1`. This means the attestations from `slot - 1` aren't taken into consideration, and any boost applied to the block from `slot - 1` is not removed (it should be). In the language of the linked spec issue, this means we are liable to attest to C, even when the majority voting weight has already caused a re-org to B. ### Why remove the call before the state advance? If we've run fork choice at the start of the slot then it has already dequeued all the attestations from the previous slot, which are the only ones eligible to influence the head in the current slot. Running fork choice again is unnecessary (unless we run it for the next slot and try to pre-empt a re-org, but I don't currently think this is a great idea). ### Performance Based on Prater testing this adds about 5-25ms of runtime to block proposal times, which are 500-1000ms on average (and spike to 5s+ sometimes due to state handling issues :cry: ). I believe this is a small enough penalty to enable it by default, with the option to disable it via the new flag `--fork-choice-before-proposal-timeout 0`. Upcoming work on block packing and state representation will also reduce block production times in general, while removing the spikes. ### Implementation Fork choice gets invoked at the start of the slot via the `per_slot_task` function called from the slot timer. It then uses a condition variable to signal to block production that fork choice has been updated. This is a bit funky, but it seems to work. One downside of the timer-based approach is that it doesn't happen automatically in most of the tests. The test added by this PR has to trigger the run manually.
2022-05-20 05:02:11 +00:00
let randao_reveal = self.sign_randao_reveal(&state, proposer_index, slot);
2019-06-16 06:24:33 +00:00
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
let (block, state) = self
2019-06-16 06:24:33 +00:00
.chain
.produce_block_on_state(
state,
None,
slot,
randao_reveal,
Some(graffiti),
ProduceBlockVerification::VerifyRandao,
)
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();
2019-06-16 06:24:33 +00:00
let signed_block = block.sign(
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
&self.validator_keypairs[proposer_index].sk,
&state.fork(),
state.genesis_validators_root(),
&self.spec,
);
2019-06-16 06:24:33 +00:00
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
(signed_block, state)
2019-06-16 06:24:33 +00:00
}
/// Useful for the `per_block_processing` tests. Creates a block, and returns the state after
/// caches are built but before the generated block is processed.
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
pub async fn make_block_return_pre_state(
&self,
mut state: BeaconState<E>,
slot: Slot,
) -> (SignedBeaconBlock<E>, BeaconState<E>) {
assert_ne!(slot, 0, "can't produce a block at slot 0");
assert!(slot >= state.slot());
complete_state_advance(&mut state, None, slot, &self.spec)
.expect("should be able to advance state to slot");
state
.build_all_caches(&self.spec)
.expect("should build caches");
let proposer_index = state.get_beacon_proposer_index(slot, &self.spec).unwrap();
// If we produce two blocks for the same slot, they hash up to the same value and
// BeaconChain errors out with `BlockIsAlreadyKnown`. Vary the graffiti so that we produce
// different blocks each time.
let graffiti = Graffiti::from(self.rng.lock().gen::<[u8; 32]>());
Run fork choice before block proposal (#3168) ## Issue Addressed Upcoming spec change https://github.com/ethereum/consensus-specs/pull/2878 ## Proposed Changes 1. Run fork choice at the start of every slot, and wait for this run to complete before proposing a block. 2. As an optimisation, also run fork choice 3/4 of the way through the slot (at 9s), _dequeueing attestations for the next slot_. 3. Remove the fork choice run from the state advance timer that occurred before advancing the state. ## Additional Info ### Block Proposal Accuracy This change makes us more likely to propose on top of the correct head in the presence of re-orgs with proposer boost in play. The main scenario that this change is designed to address is described in the linked spec issue. ### Attestation Accuracy This change _also_ makes us more likely to attest to the correct head. Currently in the case of a skipped slot at `slot` we only run fork choice 9s into `slot - 1`. This means the attestations from `slot - 1` aren't taken into consideration, and any boost applied to the block from `slot - 1` is not removed (it should be). In the language of the linked spec issue, this means we are liable to attest to C, even when the majority voting weight has already caused a re-org to B. ### Why remove the call before the state advance? If we've run fork choice at the start of the slot then it has already dequeued all the attestations from the previous slot, which are the only ones eligible to influence the head in the current slot. Running fork choice again is unnecessary (unless we run it for the next slot and try to pre-empt a re-org, but I don't currently think this is a great idea). ### Performance Based on Prater testing this adds about 5-25ms of runtime to block proposal times, which are 500-1000ms on average (and spike to 5s+ sometimes due to state handling issues :cry: ). I believe this is a small enough penalty to enable it by default, with the option to disable it via the new flag `--fork-choice-before-proposal-timeout 0`. Upcoming work on block packing and state representation will also reduce block production times in general, while removing the spikes. ### Implementation Fork choice gets invoked at the start of the slot via the `per_slot_task` function called from the slot timer. It then uses a condition variable to signal to block production that fork choice has been updated. This is a bit funky, but it seems to work. One downside of the timer-based approach is that it doesn't happen automatically in most of the tests. The test added by this PR has to trigger the run manually.
2022-05-20 05:02:11 +00:00
let randao_reveal = self.sign_randao_reveal(&state, proposer_index, slot);
let pre_state = state.clone();
let (block, state) = self
.chain
.produce_block_on_state(
state,
None,
slot,
randao_reveal,
Some(graffiti),
ProduceBlockVerification::VerifyRandao,
)
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();
let signed_block = block.sign(
&self.validator_keypairs[proposer_index].sk,
&state.fork(),
state.genesis_validators_root(),
&self.spec,
);
(signed_block, pre_state)
}
Run fork choice before block proposal (#3168) ## Issue Addressed Upcoming spec change https://github.com/ethereum/consensus-specs/pull/2878 ## Proposed Changes 1. Run fork choice at the start of every slot, and wait for this run to complete before proposing a block. 2. As an optimisation, also run fork choice 3/4 of the way through the slot (at 9s), _dequeueing attestations for the next slot_. 3. Remove the fork choice run from the state advance timer that occurred before advancing the state. ## Additional Info ### Block Proposal Accuracy This change makes us more likely to propose on top of the correct head in the presence of re-orgs with proposer boost in play. The main scenario that this change is designed to address is described in the linked spec issue. ### Attestation Accuracy This change _also_ makes us more likely to attest to the correct head. Currently in the case of a skipped slot at `slot` we only run fork choice 9s into `slot - 1`. This means the attestations from `slot - 1` aren't taken into consideration, and any boost applied to the block from `slot - 1` is not removed (it should be). In the language of the linked spec issue, this means we are liable to attest to C, even when the majority voting weight has already caused a re-org to B. ### Why remove the call before the state advance? If we've run fork choice at the start of the slot then it has already dequeued all the attestations from the previous slot, which are the only ones eligible to influence the head in the current slot. Running fork choice again is unnecessary (unless we run it for the next slot and try to pre-empt a re-org, but I don't currently think this is a great idea). ### Performance Based on Prater testing this adds about 5-25ms of runtime to block proposal times, which are 500-1000ms on average (and spike to 5s+ sometimes due to state handling issues :cry: ). I believe this is a small enough penalty to enable it by default, with the option to disable it via the new flag `--fork-choice-before-proposal-timeout 0`. Upcoming work on block packing and state representation will also reduce block production times in general, while removing the spikes. ### Implementation Fork choice gets invoked at the start of the slot via the `per_slot_task` function called from the slot timer. It then uses a condition variable to signal to block production that fork choice has been updated. This is a bit funky, but it seems to work. One downside of the timer-based approach is that it doesn't happen automatically in most of the tests. The test added by this PR has to trigger the run manually.
2022-05-20 05:02:11 +00:00
/// Create a randao reveal for a block at `slot`.
pub fn sign_randao_reveal(
&self,
state: &BeaconState<E>,
proposer_index: usize,
slot: Slot,
) -> Signature {
let epoch = slot.epoch(E::slots_per_epoch());
let domain = self.spec.get_domain(
epoch,
Domain::Randao,
&state.fork(),
state.genesis_validators_root(),
);
let message = epoch.signing_root(domain);
let sk = &self.validator_keypairs[proposer_index].sk;
sk.sign(message)
}
/// Produces an "unaggregated" attestation for the given `slot` and `index` that attests to
/// `beacon_block_root`. The provided `state` should match the `block.state_root` for the
/// `block` identified by `beacon_block_root`.
///
/// The attestation doesn't _really_ have anything about it that makes it unaggregated per say,
/// however this function is only required in the context of forming an unaggregated
/// attestation. It would be an (undetectable) violation of the protocol to create a
/// `SignedAggregateAndProof` based upon the output of this function.
///
/// This function will produce attestations to optimistic blocks, which is against the
/// specification but useful during testing.
pub fn produce_unaggregated_attestation_for_block(
&self,
slot: Slot,
index: CommitteeIndex,
beacon_block_root: Hash256,
mut state: Cow<BeaconState<E>>,
state_root: Hash256,
) -> Result<Attestation<E>, BeaconChainError> {
let epoch = slot.epoch(E::slots_per_epoch());
if state.slot() > slot {
return Err(BeaconChainError::CannotAttestToFutureState);
} else if state.current_epoch() < epoch {
let mut_state = state.to_mut();
// Only perform a "partial" state advance since we do not require the state roots to be
// accurate.
partial_state_advance(
mut_state,
Some(state_root),
epoch.start_slot(E::slots_per_epoch()),
&self.spec,
)?;
mut_state.build_committee_cache(RelativeEpoch::Current, &self.spec)?;
}
let committee_len = state.get_beacon_committee(slot, index)?.committee.len();
let target_slot = epoch.start_slot(E::slots_per_epoch());
let target_root = if state.slot() <= target_slot {
beacon_block_root
} else {
*state.get_block_root(target_slot)?
};
Ok(Attestation {
aggregation_bits: BitList::with_capacity(committee_len)?,
data: AttestationData {
slot,
index,
beacon_block_root,
source: state.current_justified_checkpoint(),
target: Checkpoint {
epoch,
root: target_root,
},
},
signature: AggregateSignature::empty(),
})
}
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
/// A list of attestations for each committee for the given slot.
2019-06-21 01:55:37 +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
/// The first layer of the Vec is organised per committee. For example, if the return value is
/// called `all_attestations`, then all attestations in `all_attestations[0]` will be for
/// committee 0, whilst all in `all_attestations[1]` will be for committee 1.
pub fn make_unaggregated_attestations(
Attestation processing (#497) * Renamed fork_choice::process_attestation_from_block * Processing attestation in fork choice * Retrieving state from store and checking signature * Looser check on beacon state validity. * Cleaned up get_attestation_state * Expanded fork choice api to provide latest validator message. * Checking if the an attestation contains a latest message * Correct process_attestation error handling. * Copy paste error in comment fixed. * Tidy ancestor iterators * Getting attestation slot via helper method * Refactored attestation creation in test utils * Revert "Refactored attestation creation in test utils" This reverts commit 4d277fe4239a7194758b18fb5c00dfe0b8231306. * Integration tests for free attestation processing * Implicit conflicts resolved. * formatting * Do first pass on Grants code * Add another attestation processing test * Tidy attestation processing * Remove old code fragment * Add non-compiling half finished changes * Simplify, fix bugs, add tests for chain iters * Remove attestation processing from op pool * Fix bug with fork choice, tidy * Fix overly restrictive check in fork choice. * Ensure committee cache is build during attn proc * Ignore unknown blocks at fork choice * Various minor fixes * Make fork choice write lock in to read lock * Remove unused method * Tidy comments * Fix attestation prod. target roots change * Fix compile error in store iters * Reject any attestation prior to finalization * Fix minor PR comments * Remove duplicated attestation finalization check * Remove awkward `let` statement
2019-08-14 00:55:24 +00:00
&self,
attesting_validators: &[usize],
Attestation processing (#497) * Renamed fork_choice::process_attestation_from_block * Processing attestation in fork choice * Retrieving state from store and checking signature * Looser check on beacon state validity. * Cleaned up get_attestation_state * Expanded fork choice api to provide latest validator message. * Checking if the an attestation contains a latest message * Correct process_attestation error handling. * Copy paste error in comment fixed. * Tidy ancestor iterators * Getting attestation slot via helper method * Refactored attestation creation in test utils * Revert "Refactored attestation creation in test utils" This reverts commit 4d277fe4239a7194758b18fb5c00dfe0b8231306. * Integration tests for free attestation processing * Implicit conflicts resolved. * formatting * Do first pass on Grants code * Add another attestation processing test * Tidy attestation processing * Remove old code fragment * Add non-compiling half finished changes * Simplify, fix bugs, add tests for chain iters * Remove attestation processing from op pool * Fix bug with fork choice, tidy * Fix overly restrictive check in fork choice. * Ensure committee cache is build during attn proc * Ignore unknown blocks at fork choice * Various minor fixes * Make fork choice write lock in to read lock * Remove unused method * Tidy comments * Fix attestation prod. target roots change * Fix compile error in store iters * Reject any attestation prior to finalization * Fix minor PR comments * Remove duplicated attestation finalization check * Remove awkward `let` statement
2019-08-14 00:55:24 +00:00
state: &BeaconState<E>,
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: Hash256,
head_block_root: SignedBeaconBlockHash,
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
attestation_slot: Slot,
) -> Vec<Vec<(Attestation<E>, SubnetId)>> {
let committee_count = state.get_committee_count_at_slot(state.slot()).unwrap();
let fork = self
.spec
.fork_at_epoch(attestation_slot.epoch(E::slots_per_epoch()));
state
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
.get_beacon_committees_at_slot(attestation_slot)
.expect("should get committees")
.iter()
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
.map(|bc| {
bc.committee
.par_iter()
.enumerate()
.filter_map(|(i, validator_index)| {
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
if !attesting_validators.contains(validator_index) {
return None;
}
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 mut attestation = self
.produce_unaggregated_attestation_for_block(
attestation_slot,
bc.index,
head_block_root.into(),
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
Cow::Borrowed(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
state_root,
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
)
.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
attestation.aggregation_bits.set(i, true).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
attestation.signature = {
let domain = self.spec.get_domain(
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
attestation.data.target.epoch,
Domain::BeaconAttester,
&fork,
state.genesis_validators_root(),
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 message = attestation.data.signing_root(domain);
let mut agg_sig = AggregateSignature::infinity();
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
agg_sig.add_assign(
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
&self.validator_keypairs[*validator_index].sk.sign(message),
);
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
agg_sig
};
let subnet_id = SubnetId::compute_subnet_for_attestation_data::<E>(
&attestation.data,
committee_count,
&self.chain.spec,
)
.unwrap();
Some((attestation, subnet_id))
})
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
.collect()
})
.collect()
}
/// A list of sync messages for the given state.
pub fn make_sync_committee_messages(
&self,
state: &BeaconState<E>,
head_block_root: Hash256,
message_slot: Slot,
relative_sync_committee: RelativeSyncCommittee,
) -> Vec<Vec<(SyncCommitteeMessage, usize)>> {
let sync_committee: Arc<SyncCommittee<E>> = match relative_sync_committee {
RelativeSyncCommittee::Current => state
.current_sync_committee()
.expect("should be called on altair beacon state")
.clone(),
RelativeSyncCommittee::Next => state
.next_sync_committee()
.expect("should be called on altair beacon state")
.clone(),
};
let fork = self
.spec
.fork_at_epoch(message_slot.epoch(E::slots_per_epoch()));
sync_committee
.pubkeys
.as_ref()
.chunks(E::sync_subcommittee_size())
.map(|subcommittee| {
subcommittee
.iter()
.enumerate()
.map(|(subcommittee_position, pubkey)| {
let validator_index = self
.chain
.validator_index(pubkey)
.expect("should find validator index")
.expect("pubkey should exist in the beacon chain");
let sync_message = SyncCommitteeMessage::new::<E>(
message_slot,
head_block_root,
validator_index as u64,
&self.validator_keypairs[validator_index].sk,
&fork,
state.genesis_validators_root(),
&self.spec,
);
(sync_message, subcommittee_position)
})
.collect()
})
.collect()
}
/// Deprecated: Use make_unaggregated_attestations() instead.
///
/// A list of attestations for each committee for the given slot.
///
/// The first layer of the Vec is organised per committee. For example, if the return value is
/// called `all_attestations`, then all attestations in `all_attestations[0]` will be for
/// committee 0, whilst all in `all_attestations[1]` will be for committee 1.
pub fn get_unaggregated_attestations(
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
&self,
attestation_strategy: &AttestationStrategy,
state: &BeaconState<E>,
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: Hash256,
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
head_block_root: Hash256,
attestation_slot: Slot,
) -> Vec<Vec<(Attestation<E>, SubnetId)>> {
let validators: Vec<usize> = match attestation_strategy {
AttestationStrategy::AllValidators => self.get_all_validators(),
AttestationStrategy::SomeValidators(vals) => vals.clone(),
};
self.make_unaggregated_attestations(
&validators,
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,
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,
head_block_root.into(),
attestation_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
pub fn make_attestations(
&self,
attesting_validators: &[usize],
state: &BeaconState<E>,
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: Hash256,
block_hash: SignedBeaconBlockHash,
slot: Slot,
) -> HarnessAttestations<E> {
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 unaggregated_attestations = self.make_unaggregated_attestations(
attesting_validators,
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
state_root,
block_hash,
slot,
);
let fork = self.spec.fork_at_epoch(slot.epoch(E::slots_per_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 aggregated_attestations: Vec<Option<SignedAggregateAndProof<E>>> =
unaggregated_attestations
.iter()
.map(|committee_attestations| {
// If there are any attestations in this committee, create an aggregate.
if let Some((attestation, _)) = committee_attestations.first() {
let bc = state
.get_beacon_committee(attestation.data.slot, attestation.data.index)
.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
// Find an aggregator if one exists. Return `None` if there are no
// aggregators.
let aggregator_index = bc
.committee
.iter()
.find(|&validator_index| {
if !attesting_validators.contains(validator_index) {
return false;
}
let selection_proof = SelectionProof::new::<E>(
slot,
&self.validator_keypairs[*validator_index].sk,
&fork,
state.genesis_validators_root(),
&self.spec,
);
selection_proof
.is_aggregator(bc.committee.len(), &self.spec)
.unwrap_or(false)
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
})
.copied()?;
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
// If the chain is able to produce an aggregate, use that. Otherwise, build an
// aggregate locally.
let aggregate = self
.chain
.get_aggregated_attestation(&attestation.data)
.unwrap()
.unwrap_or_else(|| {
committee_attestations.iter().skip(1).fold(
attestation.clone(),
|mut agg, (att, _)| {
agg.aggregate(att);
agg
},
)
});
let signed_aggregate = SignedAggregateAndProof::from_aggregate(
aggregator_index as u64,
aggregate,
None,
&self.validator_keypairs[aggregator_index].sk,
&fork,
state.genesis_validators_root(),
&self.spec,
);
Some(signed_aggregate)
} else {
None
}
})
.collect();
unaggregated_attestations
.into_iter()
.zip(aggregated_attestations)
.collect()
}
v0.12 fork choice update (#1229) * Incomplete scraps * Add progress on new fork choice impl * Further progress * First complete compiling version * Remove chain reference * Add new lmd_ghost crate * Start integrating into beacon chain * Update `milagro_bls` to new release (#1183) * Update milagro_bls to new release Signed-off-by: Kirk Baird <baird.k@outlook.com> * Tidy up fake cryptos Signed-off-by: Kirk Baird <baird.k@outlook.com> * move SecretHash to bls and put plaintext back Signed-off-by: Kirk Baird <baird.k@outlook.com> * Update state processing for v0.12 * Fix EF test runners for v0.12 * Fix some tests * Fix broken attestation verification test * More test fixes * Rough beacon chain impl working * Remove fork_choice_2 * Remove checkpoint manager * Half finished ssz impl * Add missed file * Add persistence * Tidy, fix some compile errors * Remove RwLock from ProtoArrayForkChoice * Fix store-based compile errors * Add comments, tidy * Move function out of ForkChoice struct * Start testing * More testing * Fix compile error * Tidy beacon_chain::fork_choice * Queue attestations from the current slot * Allow fork choice to handle prior-to-genesis start * Improve error granularity * Test attestation dequeuing * Process attestations during block * Store target root in fork choice * Move fork choice verification into new crate * Update tests * Consensus updates for v0.12 (#1228) * Update state processing for v0.12 * Fix EF test runners for v0.12 * Fix some tests * Fix broken attestation verification test * More test fixes * Fix typo found in review * Add `Block` struct to ProtoArray * Start fixing get_ancestor * Add rough progress on testing * Get fork choice tests working * Progress with testing * Fix partialeq impl * Move slot clock from fc_store * Improve testing * Add testing for best justified * Add clone back to SystemTimeSlotClock * Add balances test * Start adding balances cache again * Wire-in balances cache * Improve tests * Remove commented-out tests * Remove beacon_chain::ForkChoice * Rename crates * Update wider codebase to new fork_choice layout * Move advance_slot in test harness * Tidy ForkChoice::update_time * Fix verification tests * Fix compile error with iter::once * Fix fork choice tests * Ensure block attestations are processed * Fix failing beacon_chain tests * Add first invalid block check * Add finalized block check * Progress with testing, new store builder * Add fixes to get_ancestor * Fix old genesis justification test * Fix remaining fork choice tests * Change root iteration method * Move on_verified_block * Remove unused method * Start adding attestation verification tests * Add invalid ffg target test * Add target epoch test * Add queued attestation test * Remove old fork choice verification tests * Tidy, add test * Move fork choice lock drop * Rename BeaconForkChoiceStore * Add comments, tidy BeaconForkChoiceStore * Update metrics, rename fork_choice_store.rs * Remove genesis_block_root from ForkChoice * Tidy * Update fork_choice comments * Tidy, add comments * Tidy, simplify ForkChoice, fix compile issue * Tidy, removed dead file * Increase http request timeout * Fix failing rest_api test * Set HTTP timeout back to 5s * Apply fix to get_ancestor * Address Michael's comments * Fix typo * Revert "Fix broken attestation verification test" This reverts commit 722cdc903b12611de27916a57eeecfa3224f2279. Co-authored-by: Kirk Baird <baird.k@outlook.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-06-17 01:10:22 +00:00
pub fn make_sync_contributions(
&self,
state: &BeaconState<E>,
block_hash: Hash256,
slot: Slot,
relative_sync_committee: RelativeSyncCommittee,
) -> HarnessSyncContributions<E> {
let sync_messages =
self.make_sync_committee_messages(state, block_hash, slot, relative_sync_committee);
let sync_contributions: Vec<Option<SignedContributionAndProof<E>>> = sync_messages
.iter()
.enumerate()
.map(|(subnet_id, committee_messages)| {
// If there are any sync messages in this committee, create an aggregate.
if let Some((sync_message, subcommittee_position)) = committee_messages.first() {
let sync_committee: Arc<SyncCommittee<E>> = state
.current_sync_committee()
.expect("should be called on altair beacon state")
.clone();
let aggregator_index = sync_committee
.get_subcommittee_pubkeys(subnet_id)
.unwrap()
.iter()
.find_map(|pubkey| {
let validator_index = self
.chain
.validator_index(pubkey)
.expect("should find validator index")
.expect("pubkey should exist in the beacon chain");
let selection_proof = SyncSelectionProof::new::<E>(
slot,
subnet_id as u64,
&self.validator_keypairs[validator_index].sk,
&state.fork(),
state.genesis_validators_root(),
&self.spec,
);
selection_proof
.is_aggregator::<E>()
.expect("should determine aggregator")
.then(|| validator_index)
})?;
let default = SyncCommitteeContribution::from_message(
sync_message,
subnet_id as u64,
*subcommittee_position,
)
.expect("should derive sync contribution");
let aggregate = committee_messages.iter().skip(1).fold(
default,
|mut agg, (sig, position)| {
let contribution = SyncCommitteeContribution::from_message(
sig,
subnet_id as u64,
*position,
)
.expect("should derive sync contribution");
agg.aggregate(&contribution);
agg
},
);
let signed_aggregate = SignedContributionAndProof::from_aggregate(
aggregator_index as u64,
aggregate,
None,
&self.validator_keypairs[aggregator_index].sk,
&state.fork(),
state.genesis_validators_root(),
&self.spec,
);
Some(signed_aggregate)
} else {
None
}
})
.collect();
sync_messages.into_iter().zip(sync_contributions).collect()
}
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
pub fn make_attester_slashing(&self, validator_indices: Vec<u64>) -> AttesterSlashing<E> {
let mut attestation_1 = IndexedAttestation {
attesting_indices: VariableList::new(validator_indices).unwrap(),
data: AttestationData {
slot: Slot::new(0),
index: 0,
beacon_block_root: Hash256::zero(),
target: Checkpoint {
root: Hash256::zero(),
epoch: Epoch::new(0),
},
source: Checkpoint {
root: Hash256::zero(),
epoch: Epoch::new(0),
},
},
signature: AggregateSignature::infinity(),
};
let mut attestation_2 = attestation_1.clone();
attestation_2.data.index += 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
let fork = self.chain.canonical_head.cached_head().head_fork();
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
for attestation in &mut [&mut attestation_1, &mut attestation_2] {
for &i in &attestation.attesting_indices {
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 sk = &self.validator_keypairs[i as usize].sk;
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
let genesis_validators_root = self.chain.genesis_validators_root;
let domain = self.chain.spec.get_domain(
attestation.data.target.epoch,
Domain::BeaconAttester,
&fork,
genesis_validators_root,
);
let message = attestation.data.signing_root(domain);
attestation.signature.add_assign(&sk.sign(message));
}
}
AttesterSlashing {
attestation_1,
attestation_2,
}
}
pub fn make_attester_slashing_different_indices(
&self,
validator_indices_1: Vec<u64>,
validator_indices_2: Vec<u64>,
) -> AttesterSlashing<E> {
let data = AttestationData {
slot: Slot::new(0),
index: 0,
beacon_block_root: Hash256::zero(),
target: Checkpoint {
root: Hash256::zero(),
epoch: Epoch::new(0),
},
source: Checkpoint {
root: Hash256::zero(),
epoch: Epoch::new(0),
},
};
let mut attestation_1 = IndexedAttestation {
attesting_indices: VariableList::new(validator_indices_1).unwrap(),
data: data.clone(),
signature: AggregateSignature::infinity(),
};
let mut attestation_2 = IndexedAttestation {
attesting_indices: VariableList::new(validator_indices_2).unwrap(),
data,
signature: AggregateSignature::infinity(),
};
attestation_2.data.index += 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
let fork = self.chain.canonical_head.cached_head().head_fork();
for attestation in &mut [&mut attestation_1, &mut attestation_2] {
for &i in &attestation.attesting_indices {
let sk = &self.validator_keypairs[i as usize].sk;
let genesis_validators_root = self.chain.genesis_validators_root;
let domain = self.chain.spec.get_domain(
attestation.data.target.epoch,
Domain::BeaconAttester,
&fork,
genesis_validators_root,
);
let message = attestation.data.signing_root(domain);
attestation.signature.add_assign(&sk.sign(message));
}
}
AttesterSlashing {
attestation_1,
attestation_2,
}
}
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
pub fn make_proposer_slashing(&self, validator_index: u64) -> ProposerSlashing {
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 block_header_1 = self.chain.head_beacon_block().message().block_header();
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
block_header_1.proposer_index = validator_index;
let mut block_header_2 = block_header_1.clone();
block_header_2.state_root = Hash256::zero();
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 sk = &self.validator_keypairs[validator_index as usize].sk;
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 fork = self.chain.canonical_head.cached_head().head_fork();
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
let genesis_validators_root = self.chain.genesis_validators_root;
let mut signed_block_headers = vec![block_header_1, block_header_2]
.into_iter()
.map(|block_header| {
block_header.sign::<E>(sk, &fork, genesis_validators_root, &self.chain.spec)
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
})
.collect::<Vec<_>>();
ProposerSlashing {
signed_header_2: signed_block_headers.remove(1),
signed_header_1: signed_block_headers.remove(0),
}
}
pub fn make_voluntary_exit(&self, validator_index: u64, epoch: Epoch) -> SignedVoluntaryExit {
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 sk = &self.validator_keypairs[validator_index as usize].sk;
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 fork = self.chain.canonical_head.cached_head().head_fork();
Implement standard eth2.0 API (#1569) - Resolves #1550 - Resolves #824 - Resolves #825 - Resolves #1131 - Resolves #1411 - Resolves #1256 - Resolve #1177 - Includes the `ShufflingId` struct initially defined in #1492. That PR is now closed and the changes are included here, with significant bug fixes. - Implement the https://github.com/ethereum/eth2.0-APIs in a new `http_api` crate using `warp`. This replaces the `rest_api` crate. - Add a new `common/eth2` crate which provides a wrapper around `reqwest`, providing the HTTP client that is used by the validator client and for testing. This replaces the `common/remote_beacon_node` crate. - Create a `http_metrics` crate which is a dedicated server for Prometheus metrics (they are no longer served on the same port as the REST API). We now have flags for `--metrics`, `--metrics-address`, etc. - Allow the `subnet_id` to be an optional parameter for `VerifiedUnaggregatedAttestation::verify`. This means it does not need to be provided unnecessarily by the validator client. - Move `fn map_attestation_committee` in `mod beacon_chain::attestation_verification` to a new `fn with_committee_cache` on the `BeaconChain` so the same cache can be used for obtaining validator duties. - Add some other helpers to `BeaconChain` to assist with common API duties (e.g., `block_root_at_slot`, `head_beacon_block_root`). - Change the `NaiveAggregationPool` so it can index attestations by `hash_tree_root(attestation.data)`. This is a requirement of the API. - Add functions to `BeaconChainHarness` to allow it to create slashings and exits. - Allow for `eth1::Eth1NetworkId` to go to/from a `String`. - Add functions to the `OperationPool` to allow getting all objects in the pool. - Add function to `BeaconState` to check if a committee cache is initialized. - Fix bug where `seconds_per_eth1_block` was not transferring over from `YamlConfig` to `ChainSpec`. - Add the `deposit_contract_address` to `YamlConfig` and `ChainSpec`. We needed to be able to return it in an API response. - Change some uses of serde `serialize_with` and `deserialize_with` to a single use of `with` (code quality). - Impl `Display` and `FromStr` for several BLS fields. - Check for clock discrepancy when VC polls BN for sync state (with +/- 1 slot tolerance). This is not intended to be comprehensive, it was just easy to do. - See #1434 for a per-endpoint overview. - Seeking clarity here: https://github.com/ethereum/eth2.0-APIs/issues/75 - [x] Add docs for prom port to close #1256 - [x] Follow up on this #1177 - [x] ~~Follow up with #1424~~ Will fix in future PR. - [x] Follow up with #1411 - [x] ~~Follow up with #1260~~ Will fix in future PR. - [x] Add quotes to all integers. - [x] Remove `rest_types` - [x] Address missing beacon block error. (#1629) - [x] ~~Add tests for lighthouse/peers endpoints~~ Wontfix - [x] ~~Follow up with validator status proposal~~ Tracked in #1434 - [x] Unify graffiti structs - [x] ~~Start server when waiting for genesis?~~ Will fix in future PR. - [x] TODO in http_api tests - [x] Move lighthouse endpoints off /eth/v1 - [x] Update docs to link to standard - ~~Blocked on #1586~~ Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-09-29 03:46:54 +00:00
let genesis_validators_root = self.chain.genesis_validators_root;
VoluntaryExit {
epoch,
validator_index,
}
.sign(sk, &fork, genesis_validators_root, &self.chain.spec)
}
pub fn add_voluntary_exit(
&self,
block: &mut BeaconBlock<E>,
validator_index: u64,
epoch: Epoch,
) {
let exit = self.make_voluntary_exit(validator_index, epoch);
block.body_mut().voluntary_exits_mut().push(exit).unwrap();
}
/// Create a new block, apply `block_modifier` to it, sign it and return it.
///
/// The state returned is a pre-block state at the same slot as the produced 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
pub async fn make_block_with_modifier(
&self,
state: BeaconState<E>,
slot: Slot,
block_modifier: impl FnOnce(&mut BeaconBlock<E>),
) -> (SignedBeaconBlock<E>, BeaconState<E>) {
assert_ne!(slot, 0, "can't produce a block at slot 0");
assert!(slot >= 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
let (block, state) = self.make_block_return_pre_state(state, slot).await;
let (mut block, _) = block.deconstruct();
block_modifier(&mut block);
let proposer_index = state.get_beacon_proposer_index(slot, &self.spec).unwrap();
let signed_block = block.sign(
&self.validator_keypairs[proposer_index as usize].sk,
&state.fork(),
state.genesis_validators_root(),
&self.spec,
);
(signed_block, state)
}
pub fn make_deposits<'a>(
&self,
state: &'a mut BeaconState<E>,
num_deposits: usize,
invalid_pubkey: Option<PublicKeyBytes>,
invalid_signature: Option<SignatureBytes>,
) -> (Vec<Deposit>, &'a mut BeaconState<E>) {
let mut datas = vec![];
for _ in 0..num_deposits {
let keypair = Keypair::random();
let pubkeybytes = PublicKeyBytes::from(keypair.pk.clone());
let mut data = DepositData {
pubkey: pubkeybytes,
withdrawal_credentials: Hash256::from_slice(
&get_withdrawal_credentials(&keypair.pk, self.spec.bls_withdrawal_prefix_byte)
[..],
),
amount: self.spec.min_deposit_amount,
signature: SignatureBytes::empty(),
};
data.signature = data.create_signature(&keypair.sk, &self.spec);
if let Some(invalid_pubkey) = invalid_pubkey {
data.pubkey = invalid_pubkey;
}
if let Some(invalid_signature) = invalid_signature.clone() {
data.signature = invalid_signature;
}
datas.push(data);
}
// Vector containing all leaves
let leaves = datas
.iter()
.map(|data| data.tree_hash_root())
.collect::<Vec<_>>();
// Building a VarList from leaves
let deposit_data_list = VariableList::<_, U4294967296>::from(leaves.clone());
// Setting the deposit_root to be the tree_hash_root of the VarList
state.eth1_data_mut().deposit_root = deposit_data_list.tree_hash_root();
state.eth1_data_mut().deposit_count = num_deposits as u64;
*state.eth1_deposit_index_mut() = 0;
// Building the merkle tree used for generating proofs
let tree = MerkleTree::create(&leaves[..], self.spec.deposit_contract_tree_depth as usize);
// Building proofs
let mut proofs = vec![];
for i in 0..leaves.len() {
let (_, mut proof) =
tree.generate_proof(i, self.spec.deposit_contract_tree_depth as usize);
proof.push(Hash256::from_slice(&int_to_bytes32(leaves.len() as u64)));
proofs.push(proof);
}
// Building deposits
let deposits = datas
.into_par_iter()
.zip(proofs.into_par_iter())
.map(|(data, proof)| (data, proof.into()))
.map(|(data, proof)| Deposit { proof, data })
.collect::<Vec<_>>();
// Pushing deposits to block body
(deposits, 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
pub async fn process_block(
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
&self,
slot: Slot,
block: SignedBeaconBlock<E>,
) -> Result<SignedBeaconBlockHash, BlockError<E>> {
self.set_current_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
let block_hash: SignedBeaconBlockHash =
self.chain.process_block(Arc::new(block)).await?.into();
self.chain.recompute_head_at_current_slot().await?;
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
Ok(block_hash)
}
v0.12 fork choice update (#1229) * Incomplete scraps * Add progress on new fork choice impl * Further progress * First complete compiling version * Remove chain reference * Add new lmd_ghost crate * Start integrating into beacon chain * Update `milagro_bls` to new release (#1183) * Update milagro_bls to new release Signed-off-by: Kirk Baird <baird.k@outlook.com> * Tidy up fake cryptos Signed-off-by: Kirk Baird <baird.k@outlook.com> * move SecretHash to bls and put plaintext back Signed-off-by: Kirk Baird <baird.k@outlook.com> * Update state processing for v0.12 * Fix EF test runners for v0.12 * Fix some tests * Fix broken attestation verification test * More test fixes * Rough beacon chain impl working * Remove fork_choice_2 * Remove checkpoint manager * Half finished ssz impl * Add missed file * Add persistence * Tidy, fix some compile errors * Remove RwLock from ProtoArrayForkChoice * Fix store-based compile errors * Add comments, tidy * Move function out of ForkChoice struct * Start testing * More testing * Fix compile error * Tidy beacon_chain::fork_choice * Queue attestations from the current slot * Allow fork choice to handle prior-to-genesis start * Improve error granularity * Test attestation dequeuing * Process attestations during block * Store target root in fork choice * Move fork choice verification into new crate * Update tests * Consensus updates for v0.12 (#1228) * Update state processing for v0.12 * Fix EF test runners for v0.12 * Fix some tests * Fix broken attestation verification test * More test fixes * Fix typo found in review * Add `Block` struct to ProtoArray * Start fixing get_ancestor * Add rough progress on testing * Get fork choice tests working * Progress with testing * Fix partialeq impl * Move slot clock from fc_store * Improve testing * Add testing for best justified * Add clone back to SystemTimeSlotClock * Add balances test * Start adding balances cache again * Wire-in balances cache * Improve tests * Remove commented-out tests * Remove beacon_chain::ForkChoice * Rename crates * Update wider codebase to new fork_choice layout * Move advance_slot in test harness * Tidy ForkChoice::update_time * Fix verification tests * Fix compile error with iter::once * Fix fork choice tests * Ensure block attestations are processed * Fix failing beacon_chain tests * Add first invalid block check * Add finalized block check * Progress with testing, new store builder * Add fixes to get_ancestor * Fix old genesis justification test * Fix remaining fork choice tests * Change root iteration method * Move on_verified_block * Remove unused method * Start adding attestation verification tests * Add invalid ffg target test * Add target epoch test * Add queued attestation test * Remove old fork choice verification tests * Tidy, add test * Move fork choice lock drop * Rename BeaconForkChoiceStore * Add comments, tidy BeaconForkChoiceStore * Update metrics, rename fork_choice_store.rs * Remove genesis_block_root from ForkChoice * Tidy * Update fork_choice comments * Tidy, add comments * Tidy, simplify ForkChoice, fix compile issue * Tidy, removed dead file * Increase http request timeout * Fix failing rest_api test * Set HTTP timeout back to 5s * Apply fix to get_ancestor * Address Michael's comments * Fix typo * Revert "Fix broken attestation verification test" This reverts commit 722cdc903b12611de27916a57eeecfa3224f2279. Co-authored-by: Kirk Baird <baird.k@outlook.com> Co-authored-by: Michael Sproul <michael@sigmaprime.io>
2020-06-17 01:10:22 +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
pub async fn process_block_result(
&self,
block: SignedBeaconBlock<E>,
) -> Result<SignedBeaconBlockHash, BlockError<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
let block_hash: SignedBeaconBlockHash =
self.chain.process_block(Arc::new(block)).await?.into();
self.chain.recompute_head_at_current_slot().await?;
Ok(block_hash)
}
pub fn process_attestations(&self, attestations: HarnessAttestations<E>) {
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
let num_validators = self.validator_keypairs.len();
let mut unaggregated = Vec::with_capacity(num_validators);
// This is an over-allocation, but it should be fine. It won't be *that* memory hungry and
// it's nice to have fast tests.
let mut aggregated = Vec::with_capacity(num_validators);
for (unaggregated_attestations, maybe_signed_aggregate) in attestations.iter() {
for (attn, subnet) in unaggregated_attestations {
unaggregated.push((attn, Some(*subnet)));
}
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
if let Some(a) = maybe_signed_aggregate {
aggregated.push(a)
}
}
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
for result in self
.chain
.batch_verify_unaggregated_attestations_for_gossip(unaggregated.into_iter())
.unwrap()
{
let verified = result.unwrap();
self.chain.add_to_naive_aggregation_pool(&verified).unwrap();
}
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
for result in self
.chain
.batch_verify_aggregated_attestations_for_gossip(aggregated.into_iter())
.unwrap()
{
let verified = result.unwrap();
self.chain
.apply_attestation_to_fork_choice(&verified)
.unwrap();
self.chain.add_to_block_inclusion_pool(&verified).unwrap();
}
}
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
pub fn set_current_slot(&self, slot: Slot) {
let current_slot = self.chain.slot().unwrap();
let current_epoch = current_slot.epoch(E::slots_per_epoch());
let epoch = slot.epoch(E::slots_per_epoch());
assert!(
epoch >= current_epoch,
"Jumping backwards to an earlier epoch isn't well defined. \
Please generate test blocks epoch-by-epoch instead."
);
self.chain.slot_clock.set_slot(slot.into());
}
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
pub async fn add_block_at_slot(
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
&self,
slot: Slot,
state: BeaconState<E>,
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
) -> Result<(SignedBeaconBlockHash, SignedBeaconBlock<E>, BeaconState<E>), BlockError<E>> {
self.set_current_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
let (block, new_state) = self.make_block(state, slot).await;
let block_hash = self.process_block(slot, block.clone()).await?;
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
Ok((block_hash, block, new_state))
}
pub fn attest_block(
&self,
state: &BeaconState<E>,
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: Hash256,
block_hash: SignedBeaconBlockHash,
block: &SignedBeaconBlock<E>,
validators: &[usize],
) {
let attestations =
self.make_attestations(validators, state, state_root, block_hash, block.slot());
self.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
pub async fn add_attested_block_at_slot(
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
&self,
slot: Slot,
state: BeaconState<E>,
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: Hash256,
validators: &[usize],
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
) -> Result<(SignedBeaconBlockHash, BeaconState<E>), BlockError<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
let (block_hash, block, state) = self.add_block_at_slot(slot, state).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
self.attest_block(&state, state_root, block_hash, &block, validators);
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
Ok((block_hash, 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
pub async fn add_attested_blocks_at_slots(
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
&self,
state: BeaconState<E>,
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: Hash256,
slots: &[Slot],
validators: &[usize],
) -> AddBlocksResult<E> {
assert!(!slots.is_empty());
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
self.add_attested_blocks_at_slots_given_lbh(state, state_root, slots, validators, 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
.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
async fn add_attested_blocks_at_slots_given_lbh(
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
&self,
mut state: BeaconState<E>,
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: Hash256,
slots: &[Slot],
validators: &[usize],
mut latest_block_hash: Option<SignedBeaconBlockHash>,
) -> AddBlocksResult<E> {
assert!(
slots.windows(2).all(|w| w[0] <= w[1]),
"Slots have to be sorted"
); // slice.is_sorted() isn't stabilized at the moment of writing this
let mut block_hash_from_slot: HashMap<Slot, SignedBeaconBlockHash> = HashMap::new();
let mut state_hash_from_slot: HashMap<Slot, BeaconStateHash> = HashMap::new();
for slot in slots {
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 (block_hash, new_state) = self
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
.add_attested_block_at_slot(*slot, state, state_root, 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
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
.unwrap();
state = new_state;
block_hash_from_slot.insert(*slot, block_hash);
state_hash_from_slot.insert(*slot, state.tree_hash_root().into());
latest_block_hash = Some(block_hash);
}
(
block_hash_from_slot,
state_hash_from_slot,
latest_block_hash.unwrap(),
state,
)
}
/// A monstrosity of great usefulness.
///
/// Calls `add_attested_blocks_at_slots` for each of the chains in `chains`,
/// taking care to batch blocks by epoch so that the slot clock gets advanced one
/// epoch at a time.
///
/// Chains is a vec of `(state, slots, validators)` tuples.
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
pub async fn add_blocks_on_multiple_chains(
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
&self,
chains: Vec<(BeaconState<E>, Vec<Slot>, Vec<usize>)>,
) -> Vec<AddBlocksResult<E>> {
let slots_per_epoch = E::slots_per_epoch();
let min_epoch = chains
.iter()
.map(|(_, slots, _)| slots.iter().min().unwrap())
.min()
.unwrap()
.epoch(slots_per_epoch);
let max_epoch = chains
.iter()
.map(|(_, slots, _)| slots.iter().max().unwrap())
.max()
.unwrap()
.epoch(slots_per_epoch);
let mut chains = chains
.into_iter()
.map(|(state, slots, validators)| {
(
state,
slots,
validators,
HashMap::new(),
HashMap::new(),
SignedBeaconBlockHash::from(Hash256::zero()),
)
})
.collect::<Vec<_>>();
for epoch in min_epoch.as_u64()..=max_epoch.as_u64() {
let mut new_chains = vec![];
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
for (
mut head_state,
slots,
validators,
mut block_hashes,
mut state_hashes,
head_block,
) in chains
{
let epoch_slots = slots
.iter()
.filter(|s| s.epoch(slots_per_epoch).as_u64() == epoch)
.copied()
.collect::<Vec<_>>();
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 head_state_root = head_state.update_tree_hash_cache().unwrap();
let (new_block_hashes, new_state_hashes, new_head_block, new_head_state) = self
.add_attested_blocks_at_slots_given_lbh(
head_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_state_root,
&epoch_slots,
&validators,
Some(head_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
)
.await;
block_hashes.extend(new_block_hashes);
state_hashes.extend(new_state_hashes);
new_chains.push((
new_head_state,
slots,
validators,
block_hashes,
state_hashes,
new_head_block,
));
}
chains = new_chains;
}
chains
.into_iter()
.map(|(state, _, _, block_hashes, state_hashes, head_block)| {
(block_hashes, state_hashes, head_block, state)
})
.collect()
}
pub fn get_finalized_checkpoints(&self) -> HashSet<SignedBeaconBlockHash> {
let chain_dump = self.chain.chain_dump().unwrap();
chain_dump
.iter()
.cloned()
.map(|checkpoint| checkpoint.beacon_state.finalized_checkpoint().root.into())
.filter(|block_hash| *block_hash != Hash256::zero().into())
.collect()
}
/// Deprecated: Do not modify the slot clock manually; rely on add_attested_blocks_at_slots()
/// instead
///
/// Advance the slot of the `BeaconChain`.
///
/// Does not produce blocks or attestations.
pub fn advance_slot(&self) {
self.chain.slot_clock.advance_slot();
}
/// Deprecated: Use make_block() instead
///
/// Returns a newly created block, signed by the proposer for the given 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
pub async fn build_block(
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
&self,
state: BeaconState<E>,
slot: Slot,
_block_strategy: BlockStrategy,
) -> (SignedBeaconBlock<E>, BeaconState<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
self.make_block(state, slot).await
}
/// Uses `Self::extend_chain` to build the chain out to the `target_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
pub async fn extend_to_slot(&self, target_slot: Slot) -> Hash256 {
if self.chain.slot().unwrap() == self.chain.canonical_head.cached_head().head_slot() {
self.advance_slot();
}
let num_slots = target_slot
.as_usize()
.checked_sub(self.chain.slot().unwrap().as_usize())
.expect("target_slot must be >= current_slot")
.checked_add(1)
.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
self.extend_slots(num_slots).await
}
/// Uses `Self::extend_chain` to `num_slots` blocks.
///
/// Utilizes:
///
/// - BlockStrategy::OnCanonicalHead,
/// - AttestationStrategy::AllValidators,
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
pub async fn extend_slots(&self, num_slots: usize) -> Hash256 {
if self.chain.slot().unwrap() == self.chain.canonical_head.cached_head().head_slot() {
self.advance_slot();
}
self.extend_chain(
num_slots,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
)
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
}
/// Deprecated: Use add_attested_blocks_at_slots() instead
///
/// Extend the `BeaconChain` with some blocks and attestations. Returns the root of the
/// last-produced block (the head of the chain).
///
/// Chain will be extended by `num_blocks` blocks.
///
/// The `block_strategy` dictates where the new blocks will be placed.
///
/// The `attestation_strategy` dictates which validators will attest to the newly created
/// 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
pub async fn extend_chain(
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
&self,
num_blocks: usize,
block_strategy: BlockStrategy,
attestation_strategy: AttestationStrategy,
) -> Hash256 {
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, slots) = match block_strategy {
BlockStrategy::OnCanonicalHead => {
let current_slot: u64 = self.get_current_slot().into();
let slots: Vec<Slot> = (current_slot..(current_slot + (num_blocks as u64)))
.map(Slot::new)
.collect();
let state = self.get_current_state();
(state, slots)
}
BlockStrategy::ForkCanonicalChainAt {
previous_slot,
first_slot,
} => {
let first_slot_: u64 = first_slot.into();
let slots: Vec<Slot> = (first_slot_..(first_slot_ + (num_blocks as u64)))
.map(Slot::new)
.collect();
let state = self
.chain
.state_at_slot(previous_slot, StateSkipConfig::WithStateRoots)
.unwrap();
(state, slots)
}
};
let validators = match attestation_strategy {
AttestationStrategy::AllValidators => self.get_all_validators(),
AttestationStrategy::SomeValidators(vals) => vals,
};
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 (_, _, last_produced_block_hash, _) = self
.add_attested_blocks_at_slots(state, state_root, &slots, &validators)
.await;
last_produced_block_hash.into()
}
/// Deprecated: Use add_attested_blocks_at_slots() instead
///
/// Creates two forks:
///
/// - The "honest" fork: created by the `honest_validators` who have built `honest_fork_blocks`
/// on the head
/// - The "faulty" fork: created by the `faulty_validators` who skipped a slot and
/// then built `faulty_fork_blocks`.
///
/// Returns `(honest_head, faulty_head)`, the roots of the blocks at the top of each 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
pub async fn generate_two_forks_by_skipping_a_block(
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
&self,
honest_validators: &[usize],
faulty_validators: &[usize],
honest_fork_blocks: usize,
faulty_fork_blocks: usize,
) -> (Hash256, Hash256) {
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 initial_head_slot = self.chain.head_snapshot().beacon_block.slot();
// Move to the next slot so we may produce some more blocks on the head.
self.advance_slot();
// Extend the chain with blocks where only honest validators agree.
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 honest_head = self
.extend_chain(
honest_fork_blocks,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::SomeValidators(honest_validators.to_vec()),
)
.await;
// Go back to the last block where all agreed, and build blocks upon it where only faulty nodes
// agree.
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 faulty_head = self
.extend_chain(
faulty_fork_blocks,
BlockStrategy::ForkCanonicalChainAt {
previous_slot: initial_head_slot,
// `initial_head_slot + 2` means one slot is skipped.
first_slot: initial_head_slot + 2,
},
AttestationStrategy::SomeValidators(faulty_validators.to_vec()),
)
.await;
assert_ne!(honest_head, faulty_head, "forks should be distinct");
(honest_head, faulty_head)
}
2019-06-16 06:24:33 +00:00
}
Add `execution_optimistic` flag to HTTP responses (#3070) ## Issue Addressed #3031 ## Proposed Changes Updates the following API endpoints to conform with https://github.com/ethereum/beacon-APIs/pull/190 and https://github.com/ethereum/beacon-APIs/pull/196 - [x] `beacon/states/{state_id}/root` - [x] `beacon/states/{state_id}/fork` - [x] `beacon/states/{state_id}/finality_checkpoints` - [x] `beacon/states/{state_id}/validators` - [x] `beacon/states/{state_id}/validators/{validator_id}` - [x] `beacon/states/{state_id}/validator_balances` - [x] `beacon/states/{state_id}/committees` - [x] `beacon/states/{state_id}/sync_committees` - [x] `beacon/headers` - [x] `beacon/headers/{block_id}` - [x] `beacon/blocks/{block_id}` - [x] `beacon/blocks/{block_id}/root` - [x] `beacon/blocks/{block_id}/attestations` - [x] `debug/beacon/states/{state_id}` - [x] `debug/beacon/heads` - [x] `validator/duties/attester/{epoch}` - [x] `validator/duties/proposer/{epoch}` - [x] `validator/duties/sync/{epoch}` Updates the following Server-Sent Events: - [x] `events?topics=head` - [x] `events?topics=block` - [x] `events?topics=finalized_checkpoint` - [x] `events?topics=chain_reorg` ## Backwards Incompatible There is a very minor breaking change with the way the API now handles requests to `beacon/blocks/{block_id}/root` and `beacon/states/{state_id}/root` when `block_id` or `state_id` is the `Root` variant of `BlockId` and `StateId` respectively. Previously a request to a non-existent root would simply echo the root back to the requester: ``` curl "http://localhost:5052/eth/v1/beacon/states/0xaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa/root" {"data":{"root":"0xaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"}} ``` Now it will return a `404`: ``` curl "http://localhost:5052/eth/v1/beacon/blocks/0xaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa/root" {"code":404,"message":"NOT_FOUND: beacon block with root 0xaaaa…aaaa","stacktraces":[]} ``` In addition to this is the block root `0x0000000000000000000000000000000000000000000000000000000000000000` previously would return the genesis block. It will now return a `404`: ``` curl "http://localhost:5052/eth/v1/beacon/blocks/0x0000000000000000000000000000000000000000000000000000000000000000" {"code":404,"message":"NOT_FOUND: beacon block with root 0x0000…0000","stacktraces":[]} ``` ## Additional Info - `execution_optimistic` is always set, and will return `false` pre-Bellatrix. I am also open to the idea of doing something like `#[serde(skip_serializing_if = "Option::is_none")]`. - The value of `execution_optimistic` is set to `false` where possible. Any computation that is reliant on the `head` will simply use the `ExecutionStatus` of the head (unless the head block is pre-Bellatrix). Co-authored-by: Paul Hauner <paul@paulhauner.com>
2022-07-25 08:23:00 +00:00
// Junk `Debug` impl to satistfy certain trait bounds during testing.
impl<T: BeaconChainTypes> fmt::Debug for BeaconChainHarness<T> {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "BeaconChainHarness")
}
}