lighthouse/beacon_node/beacon_chain/src/canonical_head.rs

1386 lines
56 KiB
Rust
Raw Normal View History

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
//! This module provides all functionality for finding the canonical head, updating all necessary
//! components (e.g. caches) and maintaining a cached head block and state.
//!
//! For practically all applications, the "canonical head" can be read using
//! `beacon_chain.canonical_head.cached_head()`.
//!
//! The canonical head can be updated using `beacon_chain.recompute_head()`.
//!
//! ## Deadlock safety
//!
//! This module contains three locks:
//!
//! 1. `RwLock<BeaconForkChoice>`: Contains `proto_array` fork choice.
//! 2. `RwLock<CachedHead>`: Contains a cached block/state from the last run of `proto_array`.
//! 3. `Mutex<()>`: Is used to prevent concurrent execution of `BeaconChain::recompute_head`.
//!
//! This module has to take great efforts to avoid causing a deadlock with these three methods. Any
//! developers working in this module should tread carefully and seek a detailed review.
//!
//! To encourage safe use of this module, it should **only ever return a read or write lock for the
//! fork choice lock (lock 1)**. Whilst public functions might indirectly utilise locks (2) and (3),
//! the fundamental `RwLockWriteGuard` or `RwLockReadGuard` should never be exposed. This prevents
//! external functions from acquiring these locks in conflicting orders and causing a deadlock.
//!
//! ## Design Considerations
//!
//! We separate the `BeaconForkChoice` and `CachedHead` into two `RwLocks` because we want to ensure
//! fast access to the `CachedHead`. If we were to put them both under the same lock, we would need
//! to take an exclusive write-lock on it in order to run `ForkChoice::get_head`. This can take tens
//! of milliseconds and would block all downstream functions that want to know simple things like
//! the head block root. This is unacceptable for fast-responding functions like the networking
//! stack.
use crate::persisted_fork_choice::PersistedForkChoice;
use crate::{
beacon_chain::{
BeaconForkChoice, BeaconStore, BLOCK_PROCESSING_CACHE_LOCK_TIMEOUT, FORK_CHOICE_DB_KEY,
},
block_times_cache::BlockTimesCache,
events::ServerSentEventHandler,
metrics,
validator_monitor::{get_slot_delay_ms, timestamp_now},
BeaconChain, BeaconChainError as Error, BeaconChainTypes, BeaconSnapshot,
};
use eth2::types::{EventKind, SseChainReorg, SseFinalizedCheckpoint, SseHead, SseLateHead};
Reset payload statuses when resuming fork choice (#3498) ## Issue Addressed NA ## Proposed Changes This PR is motivated by a recent consensus failure in Geth where it returned `INVALID` for an `VALID` block. Without this PR, the only way to recover is by re-syncing Lighthouse. Whilst ELs "shouldn't have consensus failures", in reality it's something that we can expect from time to time due to the complex nature of Ethereum. Being able to recover easily will help the network recover and EL devs to troubleshoot. The risk introduced with this PR is that genuinely INVALID payloads get a "second chance" at being imported. I believe the DoS risk here is negligible since LH needs to be restarted in order to re-process the payload. Furthermore, there's no reason to think that a well-performing EL will accept a truly invalid payload the second-time-around. ## Additional Info This implementation has the following intricacies: 1. Instead of just resetting *invalid* payloads to optimistic, we'll also reset *valid* payloads. This is an artifact of our existing implementation. 1. We will only reset payload statuses when we detect an invalid payload present in `proto_array` - This helps save us from forgetting that all our blocks are valid in the "best case scenario" where there are no invalid blocks. 1. If we fail to revert the payload statuses we'll log a `CRIT` and just continue with a `proto_array` that *does not* have reverted payload statuses. - The code to revert statuses needs to deal with balances and proposer-boost, so it's a failure point. This is a defensive measure to avoid introducing new show-stopping bugs to LH.
2022-08-29 14:34:41 +00:00
use fork_choice::{
CountUnrealizedFull, ExecutionStatus, ForkChoiceView, ForkchoiceUpdateParameters, ProtoBlock,
ResetPayloadStatuses,
Reset payload statuses when resuming fork choice (#3498) ## Issue Addressed NA ## Proposed Changes This PR is motivated by a recent consensus failure in Geth where it returned `INVALID` for an `VALID` block. Without this PR, the only way to recover is by re-syncing Lighthouse. Whilst ELs "shouldn't have consensus failures", in reality it's something that we can expect from time to time due to the complex nature of Ethereum. Being able to recover easily will help the network recover and EL devs to troubleshoot. The risk introduced with this PR is that genuinely INVALID payloads get a "second chance" at being imported. I believe the DoS risk here is negligible since LH needs to be restarted in order to re-process the payload. Furthermore, there's no reason to think that a well-performing EL will accept a truly invalid payload the second-time-around. ## Additional Info This implementation has the following intricacies: 1. Instead of just resetting *invalid* payloads to optimistic, we'll also reset *valid* payloads. This is an artifact of our existing implementation. 1. We will only reset payload statuses when we detect an invalid payload present in `proto_array` - This helps save us from forgetting that all our blocks are valid in the "best case scenario" where there are no invalid blocks. 1. If we fail to revert the payload statuses we'll log a `CRIT` and just continue with a `proto_array` that *does not* have reverted payload statuses. - The code to revert statuses needs to deal with balances and proposer-boost, so it's a failure point. This is a defensive measure to avoid introducing new show-stopping bugs to LH.
2022-08-29 14:34:41 +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
use itertools::process_results;
use parking_lot::{Mutex, RwLock, RwLockReadGuard, RwLockWriteGuard};
use slog::{crit, debug, error, warn, Logger};
use slot_clock::SlotClock;
use std::sync::Arc;
use std::time::Duration;
use store::{iter::StateRootsIterator, KeyValueStoreOp, StoreItem};
use task_executor::{JoinHandle, ShutdownReason};
use types::*;
/// Simple wrapper around `RwLock` that uses private visibility to prevent any other modules from
/// accessing the contained lock without it being explicitly noted in this module.
pub struct CanonicalHeadRwLock<T>(RwLock<T>);
impl<T> From<RwLock<T>> for CanonicalHeadRwLock<T> {
fn from(rw_lock: RwLock<T>) -> Self {
Self(rw_lock)
}
}
impl<T> CanonicalHeadRwLock<T> {
fn new(item: T) -> Self {
Self::from(RwLock::new(item))
}
fn read(&self) -> RwLockReadGuard<T> {
self.0.read()
}
fn write(&self) -> RwLockWriteGuard<T> {
self.0.write()
}
}
/// Provides a series of cached values from the last time `BeaconChain::recompute_head` was run.
///
/// This struct is designed to be cheap-to-clone, any large fields should be wrapped in an `Arc` (or
/// similar).
#[derive(Clone)]
pub struct CachedHead<E: EthSpec> {
/// Provides the head block and state from the last time the head was updated.
pub snapshot: Arc<BeaconSnapshot<E>>,
/// The justified checkpoint as per `self.fork_choice`.
///
/// This value may be distinct to the `self.snapshot.beacon_state.justified_checkpoint`.
/// This value should be used over the beacon state value in practically all circumstances.
justified_checkpoint: Checkpoint,
/// The finalized checkpoint as per `self.fork_choice`.
///
/// This value may be distinct to the `self.snapshot.beacon_state.finalized_checkpoint`.
/// This value should be used over the beacon state value in practically all circumstances.
finalized_checkpoint: Checkpoint,
/// The `execution_payload.block_hash` of the block at the head of the chain. Set to `None`
/// before Bellatrix.
head_hash: Option<ExecutionBlockHash>,
/// The `execution_payload.block_hash` of the justified block. Set to `None` before Bellatrix.
justified_hash: Option<ExecutionBlockHash>,
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
/// The `execution_payload.block_hash` of the finalized block. Set to `None` before Bellatrix.
finalized_hash: Option<ExecutionBlockHash>,
}
impl<E: EthSpec> CachedHead<E> {
/// Returns root of the block at the head of the beacon chain.
pub fn head_block_root(&self) -> Hash256 {
self.snapshot.beacon_block_root
}
/// Returns root of the `BeaconState` at the head of the beacon chain.
///
/// ## Note
///
/// This `BeaconState` has *not* been advanced to the current slot, it has the same slot as the
/// head block.
pub fn head_state_root(&self) -> Hash256 {
self.snapshot.beacon_state_root()
}
/// Returns slot of the block at the head of the beacon chain.
///
/// ## Notes
///
/// This is *not* the current slot as per the system clock. Use `BeaconChain::slot` for the
/// system clock (aka "wall clock") slot.
pub fn head_slot(&self) -> Slot {
self.snapshot.beacon_block.slot()
}
/// Returns the `Fork` from the `BeaconState` at the head of the chain.
pub fn head_fork(&self) -> Fork {
self.snapshot.beacon_state.fork()
}
/// Returns the randao mix for the block at the head of the chain.
pub fn head_random(&self) -> Result<Hash256, BeaconStateError> {
let state = &self.snapshot.beacon_state;
let root = *state.get_randao_mix(state.current_epoch())?;
Ok(root)
}
/// Returns the active validator count for the current epoch of the head state.
///
/// Should only return `None` if the caches have not been built on the head state (this should
/// never happen).
pub fn active_validator_count(&self) -> Option<usize> {
self.snapshot
.beacon_state
.get_cached_active_validator_indices(RelativeEpoch::Current)
.map(|indices| indices.len())
.ok()
}
/// Returns the finalized checkpoint, as determined by fork choice.
///
/// ## Note
///
/// This is *not* the finalized checkpoint of the `head_snapshot.beacon_state`, rather it is the
/// best finalized checkpoint that has been observed by `self.fork_choice`. It is possible that
/// the `head_snapshot.beacon_state` finalized value is earlier than the one returned here.
pub fn finalized_checkpoint(&self) -> Checkpoint {
self.finalized_checkpoint
}
/// Returns the justified checkpoint, as determined by fork choice.
///
/// ## Note
///
/// This is *not* the "current justified checkpoint" of the `head_snapshot.beacon_state`, rather
/// it is the justified checkpoint in the view of `self.fork_choice`. It is possible that the
/// `head_snapshot.beacon_state` justified value is different to, but not conflicting with, the
/// one returned here.
pub fn justified_checkpoint(&self) -> Checkpoint {
self.justified_checkpoint
}
/// Returns the cached values of `ForkChoice::forkchoice_update_parameters`.
///
/// Useful for supplying to the execution layer.
pub fn forkchoice_update_parameters(&self) -> ForkchoiceUpdateParameters {
ForkchoiceUpdateParameters {
head_root: self.snapshot.beacon_block_root,
head_hash: self.head_hash,
justified_hash: self.justified_hash,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
finalized_hash: self.finalized_hash,
}
}
}
/// Represents the "canonical head" of the beacon chain.
///
/// The `cached_head` is elected by the `fork_choice` algorithm contained in this struct.
///
/// There is no guarantee that the state of the `fork_choice` struct will always represent the
/// `cached_head` (i.e. we may call `fork_choice` *without* updating the cached values), however
/// there is a guarantee that the `cached_head` represents some past state of `fork_choice` (i.e.
/// `fork_choice` never lags *behind* the `cached_head`).
pub struct CanonicalHead<T: BeaconChainTypes> {
/// Provides an in-memory representation of the non-finalized block tree and is used to run the
/// fork choice algorithm and determine the canonical head.
pub fork_choice: CanonicalHeadRwLock<BeaconForkChoice<T>>,
/// Provides values cached from a previous execution of `self.fork_choice.get_head`.
///
/// Although `self.fork_choice` might be slightly more advanced that this value, it is safe to
/// consider that these values represent the "canonical head" of the beacon chain.
pub cached_head: CanonicalHeadRwLock<CachedHead<T::EthSpec>>,
/// A lock used to prevent concurrent runs of `BeaconChain::recompute_head`.
///
/// This lock **should not be made public**, it should only be used inside this module.
recompute_head_lock: Mutex<()>,
}
impl<T: BeaconChainTypes> CanonicalHead<T> {
/// Instantiate `Self`.
pub fn new(
fork_choice: BeaconForkChoice<T>,
snapshot: Arc<BeaconSnapshot<T::EthSpec>>,
) -> Self {
let fork_choice_view = fork_choice.cached_fork_choice_view();
let forkchoice_update_params = fork_choice.get_forkchoice_update_parameters();
let cached_head = CachedHead {
snapshot,
justified_checkpoint: fork_choice_view.justified_checkpoint,
finalized_checkpoint: fork_choice_view.finalized_checkpoint,
head_hash: forkchoice_update_params.head_hash,
justified_hash: forkchoice_update_params.justified_hash,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
finalized_hash: forkchoice_update_params.finalized_hash,
};
Self {
fork_choice: CanonicalHeadRwLock::new(fork_choice),
cached_head: CanonicalHeadRwLock::new(cached_head),
recompute_head_lock: Mutex::new(()),
}
}
/// Load a persisted version of `BeaconForkChoice` from the `store` and restore `self` to that
/// state.
///
/// This is useful if some database corruption is expected and we wish to go back to our last
/// save-point.
pub(crate) fn restore_from_store(
&self,
// We don't actually need this value, however it's always present when we call this function
// and it needs to be dropped to prevent a dead-lock. Requiring it to be passed here is
// defensive programming.
mut fork_choice_write_lock: RwLockWriteGuard<BeaconForkChoice<T>>,
Reset payload statuses when resuming fork choice (#3498) ## Issue Addressed NA ## Proposed Changes This PR is motivated by a recent consensus failure in Geth where it returned `INVALID` for an `VALID` block. Without this PR, the only way to recover is by re-syncing Lighthouse. Whilst ELs "shouldn't have consensus failures", in reality it's something that we can expect from time to time due to the complex nature of Ethereum. Being able to recover easily will help the network recover and EL devs to troubleshoot. The risk introduced with this PR is that genuinely INVALID payloads get a "second chance" at being imported. I believe the DoS risk here is negligible since LH needs to be restarted in order to re-process the payload. Furthermore, there's no reason to think that a well-performing EL will accept a truly invalid payload the second-time-around. ## Additional Info This implementation has the following intricacies: 1. Instead of just resetting *invalid* payloads to optimistic, we'll also reset *valid* payloads. This is an artifact of our existing implementation. 1. We will only reset payload statuses when we detect an invalid payload present in `proto_array` - This helps save us from forgetting that all our blocks are valid in the "best case scenario" where there are no invalid blocks. 1. If we fail to revert the payload statuses we'll log a `CRIT` and just continue with a `proto_array` that *does not* have reverted payload statuses. - The code to revert statuses needs to deal with balances and proposer-boost, so it's a failure point. This is a defensive measure to avoid introducing new show-stopping bugs to LH.
2022-08-29 14:34:41 +00:00
reset_payload_statuses: ResetPayloadStatuses,
count_unrealized_full: CountUnrealizedFull,
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
store: &BeaconStore<T>,
spec: &ChainSpec,
Reset payload statuses when resuming fork choice (#3498) ## Issue Addressed NA ## Proposed Changes This PR is motivated by a recent consensus failure in Geth where it returned `INVALID` for an `VALID` block. Without this PR, the only way to recover is by re-syncing Lighthouse. Whilst ELs "shouldn't have consensus failures", in reality it's something that we can expect from time to time due to the complex nature of Ethereum. Being able to recover easily will help the network recover and EL devs to troubleshoot. The risk introduced with this PR is that genuinely INVALID payloads get a "second chance" at being imported. I believe the DoS risk here is negligible since LH needs to be restarted in order to re-process the payload. Furthermore, there's no reason to think that a well-performing EL will accept a truly invalid payload the second-time-around. ## Additional Info This implementation has the following intricacies: 1. Instead of just resetting *invalid* payloads to optimistic, we'll also reset *valid* payloads. This is an artifact of our existing implementation. 1. We will only reset payload statuses when we detect an invalid payload present in `proto_array` - This helps save us from forgetting that all our blocks are valid in the "best case scenario" where there are no invalid blocks. 1. If we fail to revert the payload statuses we'll log a `CRIT` and just continue with a `proto_array` that *does not* have reverted payload statuses. - The code to revert statuses needs to deal with balances and proposer-boost, so it's a failure point. This is a defensive measure to avoid introducing new show-stopping bugs to LH.
2022-08-29 14:34:41 +00:00
log: &Logger,
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
) -> Result<(), Error> {
let fork_choice = <BeaconChain<T>>::load_fork_choice(
store.clone(),
reset_payload_statuses,
count_unrealized_full,
spec,
log,
)?
.ok_or(Error::MissingPersistedForkChoice)?;
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_choice_view = fork_choice.cached_fork_choice_view();
let beacon_block_root = fork_choice_view.head_block_root;
let beacon_block = store
.get_full_block(&beacon_block_root)?
.ok_or(Error::MissingBeaconBlock(beacon_block_root))?;
let beacon_state_root = beacon_block.state_root();
let beacon_state = store
.get_state(&beacon_state_root, Some(beacon_block.slot()))?
.ok_or(Error::MissingBeaconState(beacon_state_root))?;
let snapshot = BeaconSnapshot {
beacon_block_root,
beacon_block: Arc::new(beacon_block),
beacon_state,
};
let forkchoice_update_params = fork_choice.get_forkchoice_update_parameters();
let cached_head = CachedHead {
snapshot: Arc::new(snapshot),
justified_checkpoint: fork_choice_view.justified_checkpoint,
finalized_checkpoint: fork_choice_view.finalized_checkpoint,
head_hash: forkchoice_update_params.head_hash,
justified_hash: forkchoice_update_params.justified_hash,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
finalized_hash: forkchoice_update_params.finalized_hash,
};
*fork_choice_write_lock = fork_choice;
// Avoid interleaving the fork choice and cached head locks.
drop(fork_choice_write_lock);
*self.cached_head.write() = cached_head;
Ok(())
}
/// Returns the execution status of the block at the head of the beacon chain.
///
/// This will only return `Err` in the scenario where `self.fork_choice` has advanced
/// significantly past the cached `head_snapshot`. In such a scenario it is likely prudent to
/// run `BeaconChain::recompute_head` to update the cached values.
pub fn head_execution_status(&self) -> Result<ExecutionStatus, Error> {
let head_block_root = self.cached_head().head_block_root();
self.fork_choice_read_lock()
.get_block_execution_status(&head_block_root)
.ok_or(Error::HeadMissingFromForkChoice(head_block_root))
}
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
/// Returns a clone of the `CachedHead` and the execution status of the contained head block.
///
/// This will only return `Err` in the scenario where `self.fork_choice` has advanced
/// significantly past the cached `head_snapshot`. In such a scenario it is likely prudent to
/// run `BeaconChain::recompute_head` to update the cached values.
pub fn head_and_execution_status(
&self,
) -> Result<(CachedHead<T::EthSpec>, ExecutionStatus), Error> {
let head = self.cached_head();
let head_block_root = head.head_block_root();
let execution_status = self
.fork_choice_read_lock()
.get_block_execution_status(&head_block_root)
.ok_or(Error::HeadMissingFromForkChoice(head_block_root))?;
Ok((head, execution_status))
}
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
/// Returns a clone of `self.cached_head`.
///
/// Takes a read-lock on `self.cached_head` for a short time (just long enough to clone it).
/// The `CachedHead` is designed to be fast-to-clone so this is preferred to passing back a
/// `RwLockReadGuard`, which may cause deadlock issues (see module-level documentation).
///
/// This function is safe to be public since it does not expose any locks.
pub fn cached_head(&self) -> CachedHead<T::EthSpec> {
self.cached_head_read_lock().clone()
}
/// Access a read-lock for the cached head.
///
/// This function is **not safe** to be public. See the module-level documentation for more
/// information about protecting from deadlocks.
fn cached_head_read_lock(&self) -> RwLockReadGuard<CachedHead<T::EthSpec>> {
self.cached_head.read()
}
/// Access a write-lock for the cached head.
///
/// This function is **not safe** to be public. See the module-level documentation for more
/// information about protecting from deadlocks.
fn cached_head_write_lock(&self) -> RwLockWriteGuard<CachedHead<T::EthSpec>> {
self.cached_head.write()
}
/// Access a read-lock for fork choice.
pub fn fork_choice_read_lock(&self) -> RwLockReadGuard<BeaconForkChoice<T>> {
self.fork_choice.read()
}
/// Access a write-lock for fork choice.
pub fn fork_choice_write_lock(&self) -> RwLockWriteGuard<BeaconForkChoice<T>> {
self.fork_choice.write()
}
}
impl<T: BeaconChainTypes> BeaconChain<T> {
/// Contains the "best block"; the head of the canonical `BeaconChain`.
///
/// It is important to note that the `snapshot.beacon_state` returned may not match the present slot. It
/// is the state as it was when the head block was received, which could be some slots prior to
/// now.
pub fn head(&self) -> CachedHead<T::EthSpec> {
self.canonical_head.cached_head()
}
/// Apply a function to an `Arc`-clone of the canonical head snapshot.
///
/// This method is a relic from an old implementation where the canonical head was not behind
/// an `Arc` and the canonical head lock had to be held whenever it was read. This method is
/// fine to be left here, it just seems a bit weird.
pub fn with_head<U, E>(
&self,
f: impl FnOnce(&BeaconSnapshot<T::EthSpec>) -> Result<U, E>,
) -> Result<U, E>
where
E: From<Error>,
{
let head_snapshot = self.head_snapshot();
f(&head_snapshot)
}
/// Returns the beacon block root at the head of the canonical chain.
///
/// See `Self::head` for more information.
pub fn head_beacon_block_root(&self) -> Hash256 {
self.canonical_head
.cached_head_read_lock()
.snapshot
.beacon_block_root
}
/// Returns the slot of the highest block in the canonical chain.
pub fn best_slot(&self) -> Slot {
self.canonical_head
.cached_head_read_lock()
.snapshot
.beacon_block
.slot()
}
/// Returns a `Arc` of the `BeaconSnapshot` at the head of the canonical chain.
///
/// See `Self::head` for more information.
pub fn head_snapshot(&self) -> Arc<BeaconSnapshot<T::EthSpec>> {
self.canonical_head.cached_head_read_lock().snapshot.clone()
}
/// Returns the beacon block at the head of the canonical chain.
///
/// See `Self::head` for more information.
pub fn head_beacon_block(&self) -> Arc<SignedBeaconBlock<T::EthSpec>> {
self.canonical_head
.cached_head_read_lock()
.snapshot
.beacon_block
.clone()
}
/// Returns a clone of the beacon state at the head of the canonical chain.
///
/// Cloning the head state is expensive and should generally be avoided outside of tests.
///
/// See `Self::head` for more information.
pub fn head_beacon_state_cloned(&self) -> BeaconState<T::EthSpec> {
// Don't clone whilst holding the read-lock, take an Arc-clone to reduce lock contention.
let snapshot: Arc<_> = self.head_snapshot();
snapshot
.beacon_state
.clone_with(CloneConfig::committee_caches_only())
}
/// Execute the fork choice algorithm and enthrone the result as the canonical head.
///
/// This method replaces the old `BeaconChain::fork_choice` method.
pub async fn recompute_head_at_current_slot(self: &Arc<Self>) {
match self.slot() {
Ok(current_slot) => self.recompute_head_at_slot(current_slot).await,
Err(e) => error!(
self.log,
"No slot when recomputing head";
"error" => ?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
}
/// Execute the fork choice algorithm and enthrone the result as the canonical head.
///
/// The `current_slot` is specified rather than relying on the wall-clock slot. Using a
/// different slot to the wall-clock can be useful for pushing fork choice into the next slot
/// *just* before the start of the slot. This ensures that block production can use the correct
/// head value without being delayed.
///
/// This function purposefully does *not* return a `Result`. It's possible for fork choice to
/// fail to update if there is only one viable head and it has an invalid execution payload. In
/// such a case it's critical that the `BeaconChain` keeps importing blocks so that the
/// situation can be rectified. We avoid returning an error here so that calling functions
/// can't abort block import because an error is returned here.
pub async fn recompute_head_at_slot(self: &Arc<Self>, 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
metrics::inc_counter(&metrics::FORK_CHOICE_REQUESTS);
let _timer = metrics::start_timer(&metrics::FORK_CHOICE_TIMES);
let chain = self.clone();
match self
.spawn_blocking_handle(
move || chain.recompute_head_at_slot_internal(current_slot),
"recompute_head_internal",
)
.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
{
// Fork choice returned successfully and did not need to update the EL.
Ok(Ok(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
// Fork choice returned successfully and needed to update the EL. It has returned a
// join-handle from when it spawned some async tasks. We should await those tasks.
Ok(Ok(Some(join_handle))) => match join_handle.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
// The async task completed successfully.
Ok(Some(())) => (),
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
// The async task did not complete successfully since the runtime is shutting down.
Ok(None) => {
debug!(
self.log,
"Did not update EL fork choice";
"info" => "shutting down"
);
}
// The async task did not complete successfully, tokio returned an error.
Err(e) => {
error!(
self.log,
"Did not update EL fork choice";
"error" => ?e
);
}
},
// There was an error recomputing the head.
Ok(Err(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
metrics::inc_counter(&metrics::FORK_CHOICE_ERRORS);
error!(
self.log,
"Error whist recomputing head";
"error" => ?e
);
}
// There was an error spawning the task.
Err(e) => {
error!(
self.log,
"Failed to spawn recompute head task";
"error" => ?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
}
}
}
/// A non-async (blocking) function which recomputes the canonical head and spawns async tasks.
///
/// This function performs long-running, heavy-lifting tasks which should not be performed on
/// the core `tokio` executor.
fn recompute_head_at_slot_internal(
self: &Arc<Self>,
current_slot: Slot,
) -> Result<Option<JoinHandle<Option<()>>>, Error> {
let recompute_head_lock = self.canonical_head.recompute_head_lock.lock();
// Take a clone of the current ("old") head.
let old_cached_head = self.canonical_head.cached_head();
// Determine the current ("old") fork choice parameters.
//
// It is important to read the `fork_choice_view` from the cached head rather than from fork
// choice, since the fork choice value might have changed between calls to this function. We
// are interested in the changes since we last cached the head values, not since fork choice
// was last run.
let old_view = ForkChoiceView {
head_block_root: old_cached_head.head_block_root(),
justified_checkpoint: old_cached_head.justified_checkpoint(),
finalized_checkpoint: old_cached_head.finalized_checkpoint(),
};
let mut fork_choice_write_lock = self.canonical_head.fork_choice_write_lock();
// Recompute the current head via the fork choice algorithm.
fork_choice_write_lock.get_head(current_slot, &self.spec)?;
// Downgrade the fork choice write-lock to a read lock, without allowing access to any
// other writers.
let fork_choice_read_lock = RwLockWriteGuard::downgrade(fork_choice_write_lock);
// Read the current head value from the fork choice algorithm.
let new_view = fork_choice_read_lock.cached_fork_choice_view();
// Check to ensure that the finalized block hasn't been marked as invalid. If it has,
// shut down Lighthouse.
let finalized_proto_block = fork_choice_read_lock.get_finalized_block()?;
check_finalized_payload_validity(self, &finalized_proto_block)?;
// Sanity check the finalized checkpoint.
//
// The new finalized checkpoint must be either equal to or better than the previous
// finalized checkpoint.
check_against_finality_reversion(&old_view, &new_view)?;
let new_head_proto_block = fork_choice_read_lock
.get_block(&new_view.head_block_root)
.ok_or(Error::HeadBlockMissingFromForkChoice(
new_view.head_block_root,
))?;
// Do not allow an invalid block to become the head.
//
// This check avoids the following infinite loop:
//
// 1. A new block is set as the head.
// 2. The EL is updated with the new head, and returns INVALID.
// 3. We call `process_invalid_execution_payload` and it calls this function.
// 4. This function elects an invalid block as the head.
// 5. GOTO 2
//
// In theory, fork choice should never select an invalid head (i.e., step #3 is impossible).
// However, this check is cheap.
if new_head_proto_block.execution_status.is_invalid() {
return Err(Error::HeadHasInvalidPayload {
block_root: new_head_proto_block.root,
execution_status: new_head_proto_block.execution_status,
});
}
// Exit early if the head or justified/finalized checkpoints have not changed, there's
// nothing to do.
if new_view == old_view {
debug!(
self.log,
"No change in canonical head";
"head" => ?new_view.head_block_root
);
return Ok(None);
}
// Get the parameters to update the execution layer since either the head or some finality
// parameters have changed.
let new_forkchoice_update_parameters =
fork_choice_read_lock.get_forkchoice_update_parameters();
perform_debug_logging::<T>(&old_view, &new_view, &fork_choice_read_lock, &self.log);
// Drop the read lock, it's no longer required and holding it any longer than necessary
// will just cause lock contention.
drop(fork_choice_read_lock);
// If the head has changed, update `self.canonical_head`.
let new_cached_head = if new_view.head_block_root != old_view.head_block_root {
metrics::inc_counter(&metrics::FORK_CHOICE_CHANGED_HEAD);
// Try and obtain the snapshot for `beacon_block_root` from the snapshot cache, falling
// back to a database read if that fails.
let new_snapshot = self
.snapshot_cache
.try_read_for(BLOCK_PROCESSING_CACHE_LOCK_TIMEOUT)
.and_then(|snapshot_cache| {
snapshot_cache.get_cloned(
new_view.head_block_root,
CloneConfig::committee_caches_only(),
)
})
.map::<Result<_, Error>, _>(Ok)
.unwrap_or_else(|| {
let beacon_block = self
.store
.get_full_block(&new_view.head_block_root)?
.ok_or(Error::MissingBeaconBlock(new_view.head_block_root))?;
let beacon_state_root = beacon_block.state_root();
let beacon_state: BeaconState<T::EthSpec> = self
.get_state(&beacon_state_root, Some(beacon_block.slot()))?
.ok_or(Error::MissingBeaconState(beacon_state_root))?;
Ok(BeaconSnapshot {
beacon_block: Arc::new(beacon_block),
beacon_block_root: new_view.head_block_root,
beacon_state,
})
})
.and_then(|mut snapshot| {
// Regardless of where we got the state from, attempt to build the committee
// caches.
snapshot
.beacon_state
.build_all_committee_caches(&self.spec)
.map_err(Into::into)
.map(|()| snapshot)
})?;
let new_cached_head = CachedHead {
snapshot: Arc::new(new_snapshot),
justified_checkpoint: new_view.justified_checkpoint,
finalized_checkpoint: new_view.finalized_checkpoint,
head_hash: new_forkchoice_update_parameters.head_hash,
justified_hash: new_forkchoice_update_parameters.justified_hash,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
finalized_hash: new_forkchoice_update_parameters.finalized_hash,
};
let new_head = {
// Now the new snapshot has been obtained, take a write-lock on the cached head so
// we can update it quickly.
let mut cached_head_write_lock = self.canonical_head.cached_head_write_lock();
// Enshrine the new head as the canonical cached head.
*cached_head_write_lock = new_cached_head;
// Take a clone of the cached head for later use. It is cloned whilst
// holding the write-lock to ensure we get exactly the head we just enshrined.
cached_head_write_lock.clone()
};
// Clear the early attester cache in case it conflicts with `self.canonical_head`.
self.early_attester_cache.clear();
new_head
} else {
let new_cached_head = CachedHead {
// The head hasn't changed, take a relatively cheap `Arc`-clone of the existing
// head.
snapshot: old_cached_head.snapshot.clone(),
justified_checkpoint: new_view.justified_checkpoint,
finalized_checkpoint: new_view.finalized_checkpoint,
head_hash: new_forkchoice_update_parameters.head_hash,
justified_hash: new_forkchoice_update_parameters.justified_hash,
Use async code when interacting with EL (#3244) ## Overview This rather extensive PR achieves two primary goals: 1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state. 2. Refactors fork choice, block production and block processing to `async` functions. Additionally, it achieves: - Concurrent forkchoice updates to the EL and cache pruning after a new head is selected. - Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production. - Concurrent per-block-processing and execution payload verification during block processing. - The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?): - I had to do this to deal with sending blocks into spawned tasks. - Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones. - We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap. - Avoids cloning *all the blocks* in *every chain segment* during sync. - It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough :sweat_smile:) - The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs. For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273 ## Changes to `canonical_head` and `fork_choice` Previously, the `BeaconChain` had two separate fields: ``` canonical_head: RwLock<Snapshot>, fork_choice: RwLock<BeaconForkChoice> ``` Now, we have grouped these values under a single struct: ``` canonical_head: CanonicalHead { cached_head: RwLock<Arc<Snapshot>>, fork_choice: RwLock<BeaconForkChoice> } ``` Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously. ## Breaking Changes ### The `state` (root) field in the `finalized_checkpoint` SSE event Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event: 1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`. 4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots. Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java#L171-L182) it uses [`getStateRootFromBlockRoot`](https://github.com/ConsenSys/teku/blob/de2b2801c89ef5abf983d6bf37867c37fc47121f/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java#L336-L341) which uses (1). I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku. ## Notes for Reviewers I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct. I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking". I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it. I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around. Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2. You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests: - Changing tests to be `tokio::async` tests. - Adding `.await` to fork choice, block processing and block production functions. - Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`. - Wrapping `SignedBeaconBlock` in an `Arc`. - In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant. I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic. Co-authored-by: Mac L <mjladson@pm.me>
2022-07-03 05:36:50 +00:00
finalized_hash: new_forkchoice_update_parameters.finalized_hash,
};
let mut cached_head_write_lock = self.canonical_head.cached_head_write_lock();
// Enshrine the new head as the canonical cached head. Whilst the head block hasn't
// changed, the FFG checkpoints must have changed.
*cached_head_write_lock = new_cached_head;
// Take a clone of the cached head for later use. It is cloned whilst
// holding the write-lock to ensure we get exactly the head we just enshrined.
cached_head_write_lock.clone()
};
// Alias for readability.
let new_snapshot = &new_cached_head.snapshot;
let old_snapshot = &old_cached_head.snapshot;
// If the head changed, perform some updates.
if new_snapshot.beacon_block_root != old_snapshot.beacon_block_root {
if let Err(e) =
self.after_new_head(&old_cached_head, &new_cached_head, new_head_proto_block)
{
crit!(
self.log,
"Error updating canonical head";
"error" => ?e
);
}
}
// Drop the old cache head nice and early to try and free the memory as soon as possible.
drop(old_cached_head);
// If the finalized checkpoint changed, perform some updates.
//
// The `after_finalization` function will take a write-lock on `fork_choice`, therefore it
// is a dead-lock risk to hold any other lock on fork choice at this point.
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
if new_view.finalized_checkpoint != old_view.finalized_checkpoint {
if let Err(e) =
self.after_finalization(&new_cached_head, new_view, finalized_proto_block)
{
crit!(
self.log,
"Error updating finalization";
"error" => ?e
);
}
}
// The execution layer updates might attempt to take a write-lock on fork choice, so it's
// important to ensure the fork-choice lock isn't being held.
let el_update_handle =
spawn_execution_layer_updates(self.clone(), new_forkchoice_update_parameters)?;
// We have completed recomputing the head and it's now valid for another process to do the
// same.
drop(recompute_head_lock);
Ok(Some(el_update_handle))
}
/// Perform updates to caches and other components after the canonical head has been changed.
fn after_new_head(
self: &Arc<Self>,
old_cached_head: &CachedHead<T::EthSpec>,
new_cached_head: &CachedHead<T::EthSpec>,
new_head_proto_block: ProtoBlock,
) -> Result<(), Error> {
let old_snapshot = &old_cached_head.snapshot;
let new_snapshot = &new_cached_head.snapshot;
let new_head_is_optimistic = new_head_proto_block
.execution_status
.is_optimistic_or_invalid();
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
// Detect and potentially report any re-orgs.
let reorg_distance = detect_reorg(
&old_snapshot.beacon_state,
old_snapshot.beacon_block_root,
&new_snapshot.beacon_state,
new_snapshot.beacon_block_root,
&self.spec,
&self.log,
);
// Determine if the new head is in a later epoch to the previous head.
let is_epoch_transition = old_snapshot
.beacon_block
.slot()
.epoch(T::EthSpec::slots_per_epoch())
< new_snapshot
.beacon_state
.slot()
.epoch(T::EthSpec::slots_per_epoch());
// These fields are used for server-sent events.
let state_root = new_snapshot.beacon_state_root();
let head_slot = new_snapshot.beacon_state.slot();
let dependent_root = new_snapshot
.beacon_state
.proposer_shuffling_decision_root(self.genesis_block_root);
let prev_dependent_root = new_snapshot
.beacon_state
.attester_shuffling_decision_root(self.genesis_block_root, RelativeEpoch::Current);
// Update the snapshot cache with the latest head value.
//
// This *could* be done inside `recompute_head`, however updating the head on the snapshot
// cache is not critical so we avoid placing it on a critical path. Note that this function
// will not return an error if the update fails, it will just log an error.
self.snapshot_cache
.try_write_for(BLOCK_PROCESSING_CACHE_LOCK_TIMEOUT)
.map(|mut snapshot_cache| {
snapshot_cache.update_head(new_snapshot.beacon_block_root);
})
.unwrap_or_else(|| {
error!(
self.log,
"Failed to obtain cache write lock";
"lock" => "snapshot_cache",
"task" => "update head"
);
});
observe_head_block_delays(
&mut self.block_times_cache.write(),
&new_head_proto_block,
new_snapshot.beacon_block.message().proposer_index(),
new_snapshot
.beacon_block
.message()
.body()
.graffiti()
.as_utf8_lossy(),
&self.slot_clock,
self.event_handler.as_ref(),
&self.log,
);
if is_epoch_transition || reorg_distance.is_some() {
self.persist_head_and_fork_choice()?;
self.op_pool.prune_attestations(self.epoch()?);
}
// Register server-sent-events for a new head.
if let Some(event_handler) = self
.event_handler
.as_ref()
.filter(|handler| handler.has_head_subscribers())
{
match (dependent_root, prev_dependent_root) {
(Ok(current_duty_dependent_root), Ok(previous_duty_dependent_root)) => {
event_handler.register(EventKind::Head(SseHead {
slot: head_slot,
block: new_snapshot.beacon_block_root,
state: state_root,
current_duty_dependent_root,
previous_duty_dependent_root,
epoch_transition: is_epoch_transition,
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
execution_optimistic: new_head_is_optimistic,
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
}));
}
(Err(e), _) | (_, Err(e)) => {
warn!(
self.log,
"Unable to find dependent roots, cannot register head event";
"error" => ?e
);
}
}
}
// Register a server-sent-event for a reorg (if necessary).
if let Some(depth) = reorg_distance {
if let Some(event_handler) = self
.event_handler
.as_ref()
.filter(|handler| handler.has_reorg_subscribers())
{
event_handler.register(EventKind::ChainReorg(SseChainReorg {
slot: head_slot,
depth: depth.as_u64(),
old_head_block: old_snapshot.beacon_block_root,
old_head_state: old_snapshot.beacon_state_root(),
new_head_block: new_snapshot.beacon_block_root,
new_head_state: new_snapshot.beacon_state_root(),
epoch: head_slot.epoch(T::EthSpec::slots_per_epoch()),
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
execution_optimistic: new_head_is_optimistic,
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
}));
}
}
Ok(())
}
/// Perform updates to caches and other components after the finalized checkpoint has been
/// changed.
///
/// This function will take a write-lock on `canonical_head.fork_choice`, therefore it would be
/// unwise to hold any lock on fork choice while calling this function.
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
fn after_finalization(
self: &Arc<Self>,
new_cached_head: &CachedHead<T::EthSpec>,
new_view: ForkChoiceView,
finalized_proto_block: ProtoBlock,
) -> Result<(), Error> {
let new_snapshot = &new_cached_head.snapshot;
let finalized_block_is_optimistic = finalized_proto_block
.execution_status
.is_optimistic_or_invalid();
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.op_pool
.prune_all(&new_snapshot.beacon_state, self.epoch()?);
self.observed_block_producers.write().prune(
new_view
.finalized_checkpoint
.epoch
.start_slot(T::EthSpec::slots_per_epoch()),
);
self.snapshot_cache
.try_write_for(BLOCK_PROCESSING_CACHE_LOCK_TIMEOUT)
.map(|mut snapshot_cache| {
snapshot_cache.prune(new_view.finalized_checkpoint.epoch);
debug!(
self.log,
"Snapshot cache pruned";
"new_len" => snapshot_cache.len(),
"remaining_roots" => ?snapshot_cache.beacon_block_roots(),
);
})
.unwrap_or_else(|| {
error!(
self.log,
"Failed to obtain cache write lock";
"lock" => "snapshot_cache",
"task" => "prune"
);
});
self.attester_cache
.prune_below(new_view.finalized_checkpoint.epoch);
if let Some(event_handler) = self.event_handler.as_ref() {
if event_handler.has_finalized_subscribers() {
event_handler.register(EventKind::FinalizedCheckpoint(SseFinalizedCheckpoint {
epoch: new_view.finalized_checkpoint.epoch,
block: new_view.finalized_checkpoint.root,
// Provide the state root of the latest finalized block, rather than the
// specific state root at the first slot of the finalized epoch (which
// might be a skip slot).
state: finalized_proto_block.state_root,
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
execution_optimistic: finalized_block_is_optimistic,
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
}));
}
}
// The store migration task requires the *state at the slot of the finalized epoch*,
// rather than the state of the latest finalized block. These two values will only
// differ when the first slot of the finalized epoch is a skip slot.
//
// Use the `StateRootsIterator` directly rather than `BeaconChain::state_root_at_slot`
// to ensure we use the same state that we just set as the head.
let new_finalized_slot = new_view
.finalized_checkpoint
.epoch
.start_slot(T::EthSpec::slots_per_epoch());
let new_finalized_state_root = process_results(
StateRootsIterator::new(&self.store, &new_snapshot.beacon_state),
|mut iter| {
iter.find_map(|(state_root, slot)| {
if slot == new_finalized_slot {
Some(state_root)
} else {
None
}
})
},
)?
.ok_or(Error::MissingFinalizedStateRoot(new_finalized_slot))?;
self.store_migrator.process_finalization(
new_finalized_state_root.into(),
new_view.finalized_checkpoint,
self.head_tracker.clone(),
)?;
// Take a write-lock on the canonical head and signal for it to prune.
self.canonical_head.fork_choice_write_lock().prune()?;
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
Ok(())
}
/// Return a database operation for writing fork choice to disk.
pub fn persist_fork_choice_in_batch(&self) -> KeyValueStoreOp {
Self::persist_fork_choice_in_batch_standalone(&self.canonical_head.fork_choice_read_lock())
}
/// Return a database operation for writing fork choice to disk.
pub fn persist_fork_choice_in_batch_standalone(
fork_choice: &BeaconForkChoice<T>,
) -> KeyValueStoreOp {
let persisted_fork_choice = PersistedForkChoice {
fork_choice: fork_choice.to_persisted(),
fork_choice_store: fork_choice.fc_store().to_persisted(),
};
persisted_fork_choice.as_kv_store_op(FORK_CHOICE_DB_KEY)
}
}
/// Check to see if the `finalized_proto_block` has an invalid execution payload. If so, shut down
/// Lighthouse.
///
/// ## Notes
///
/// This function is called whilst holding a write-lock on the `canonical_head`. To ensure dead-lock
/// safety, **do not take any other locks inside this function**.
fn check_finalized_payload_validity<T: BeaconChainTypes>(
chain: &BeaconChain<T>,
finalized_proto_block: &ProtoBlock,
) -> Result<(), Error> {
if let ExecutionStatus::Invalid(block_hash) = finalized_proto_block.execution_status {
crit!(
chain.log,
"Finalized block has an invalid payload";
"msg" => "You must use the `--purge-db` flag to clear the database and restart sync. \
You may be on a hostile network.",
"block_hash" => ?block_hash
);
let mut shutdown_sender = chain.shutdown_sender();
shutdown_sender
.try_send(ShutdownReason::Failure(
"Finalized block has an invalid execution payload.",
))
.map_err(Error::InvalidFinalizedPayloadShutdownError)?;
// Exit now, the node is in an invalid state.
return Err(Error::InvalidFinalizedPayload {
finalized_root: finalized_proto_block.root,
execution_block_hash: block_hash,
});
}
Ok(())
}
/// Check to ensure that the transition from `old_view` to `new_view` will not revert finality.
fn check_against_finality_reversion(
old_view: &ForkChoiceView,
new_view: &ForkChoiceView,
) -> Result<(), Error> {
let finalization_equal = new_view.finalized_checkpoint == old_view.finalized_checkpoint;
let finalization_advanced =
new_view.finalized_checkpoint.epoch > old_view.finalized_checkpoint.epoch;
if finalization_equal || finalization_advanced {
Ok(())
} else {
Err(Error::RevertedFinalizedEpoch {
old: old_view.finalized_checkpoint,
new: new_view.finalized_checkpoint,
})
}
}
fn perform_debug_logging<T: BeaconChainTypes>(
old_view: &ForkChoiceView,
new_view: &ForkChoiceView,
fork_choice: &BeaconForkChoice<T>,
log: &Logger,
) {
if new_view.head_block_root != old_view.head_block_root {
debug!(
log,
"Fork choice updated head";
"new_head_weight" => ?fork_choice
.get_block_weight(&new_view.head_block_root),
"new_head" => ?new_view.head_block_root,
"old_head_weight" => ?fork_choice
.get_block_weight(&old_view.head_block_root),
"old_head" => ?old_view.head_block_root,
)
}
if new_view.justified_checkpoint != old_view.justified_checkpoint {
debug!(
log,
"Fork choice justified";
"new_root" => ?new_view.justified_checkpoint.root,
"new_epoch" => new_view.justified_checkpoint.epoch,
"old_root" => ?old_view.justified_checkpoint.root,
"old_epoch" => old_view.justified_checkpoint.epoch,
)
}
if new_view.finalized_checkpoint != old_view.finalized_checkpoint {
debug!(
log,
"Fork choice finalized";
"new_root" => ?new_view.finalized_checkpoint.root,
"new_epoch" => new_view.finalized_checkpoint.epoch,
"old_root" => ?old_view.finalized_checkpoint.root,
"old_epoch" => old_view.finalized_checkpoint.epoch,
)
}
}
fn spawn_execution_layer_updates<T: BeaconChainTypes>(
chain: Arc<BeaconChain<T>>,
forkchoice_update_params: ForkchoiceUpdateParameters,
) -> Result<JoinHandle<Option<()>>, Error> {
let current_slot = chain
.slot_clock
.now_or_genesis()
.ok_or(Error::UnableToReadSlot)?;
chain
.task_executor
.clone()
.spawn_handle(
async move {
// Avoids raising an error before Bellatrix.
//
// See `Self::prepare_beacon_proposer` for more detail.
if chain.slot_is_prior_to_bellatrix(current_slot + 1) {
return;
}
if let Err(e) = chain
.update_execution_engine_forkchoice(current_slot, forkchoice_update_params)
.await
{
crit!(
chain.log,
"Failed to update execution head";
"error" => ?e
);
}
// Update the mechanism for preparing for block production on the execution layer.
//
// Performing this call immediately after `update_execution_engine_forkchoice_blocking`
// might result in two calls to fork choice updated, one *without* payload attributes and
// then a second *with* payload attributes.
//
// This seems OK. It's not a significant waste of EL<>CL bandwidth or resources, as far as I
// know.
if let Err(e) = chain.prepare_beacon_proposer(current_slot).await {
crit!(
chain.log,
"Failed to prepare proposers after fork choice";
"error" => ?e
);
}
},
"update_el_forkchoice",
)
.ok_or(Error::RuntimeShutdown)
}
/// Attempt to detect if the new head is not on the same chain as the previous block
/// (i.e., a re-org).
///
/// Note: this will declare a re-org if we skip `SLOTS_PER_HISTORICAL_ROOT` blocks
/// between calls to fork choice without swapping between chains. This seems like an
/// extreme-enough scenario that a warning is fine.
fn detect_reorg<E: EthSpec>(
old_state: &BeaconState<E>,
old_block_root: Hash256,
new_state: &BeaconState<E>,
new_block_root: Hash256,
spec: &ChainSpec,
log: &Logger,
) -> Option<Slot> {
let is_reorg = new_state
.get_block_root(old_state.slot())
.map_or(true, |root| *root != old_block_root);
if is_reorg {
let reorg_distance =
match find_reorg_slot(old_state, old_block_root, new_state, new_block_root, spec) {
Ok(slot) => old_state.slot().saturating_sub(slot),
Err(e) => {
warn!(
log,
"Could not find re-org depth";
"error" => format!("{:?}", e),
);
return None;
}
};
metrics::inc_counter(&metrics::FORK_CHOICE_REORG_COUNT);
metrics::inc_counter(&metrics::FORK_CHOICE_REORG_COUNT_INTEROP);
warn!(
log,
"Beacon chain re-org";
"previous_head" => ?old_block_root,
"previous_slot" => old_state.slot(),
"new_head" => ?new_block_root,
"new_slot" => new_state.slot(),
"reorg_distance" => reorg_distance,
);
Some(reorg_distance)
} else {
None
}
}
/// Iterate through the current chain to find the slot intersecting with the given beacon state.
/// The maximum depth this will search is `SLOTS_PER_HISTORICAL_ROOT`, and if that depth is reached
/// and no intersection is found, the finalized slot will be returned.
pub fn find_reorg_slot<E: EthSpec>(
old_state: &BeaconState<E>,
old_block_root: Hash256,
new_state: &BeaconState<E>,
new_block_root: Hash256,
spec: &ChainSpec,
) -> Result<Slot, Error> {
// The earliest slot for which the two chains may have a common history.
let lowest_slot = std::cmp::min(new_state.slot(), old_state.slot());
// Create an iterator across `$state`, assuming that the block at `$state.slot` has the
// block root of `$block_root`.
//
// The iterator will be skipped until the next value returns `lowest_slot`.
//
// This is a macro instead of a function or closure due to the complex types invloved
// in all the iterator wrapping.
macro_rules! aligned_roots_iter {
($state: ident, $block_root: ident) => {
std::iter::once(Ok(($state.slot(), $block_root)))
.chain($state.rev_iter_block_roots(spec))
.skip_while(|result| {
result
.as_ref()
.map_or(false, |(slot, _)| *slot > lowest_slot)
})
};
}
// Create iterators across old/new roots where iterators both start at the same slot.
let mut new_roots = aligned_roots_iter!(new_state, new_block_root);
let mut old_roots = aligned_roots_iter!(old_state, old_block_root);
// Whilst *both* of the iterators are still returning values, try and find a common
// ancestor between them.
while let (Some(old), Some(new)) = (old_roots.next(), new_roots.next()) {
let (old_slot, old_root) = old?;
let (new_slot, new_root) = new?;
// Sanity check to detect programming errors.
if old_slot != new_slot {
return Err(Error::InvalidReorgSlotIter { new_slot, old_slot });
}
if old_root == new_root {
// A common ancestor has been found.
return Ok(old_slot);
}
}
// If no common ancestor is found, declare that the re-org happened at the previous
// finalized slot.
//
// Sometimes this will result in the return slot being *lower* than the actual reorg
// slot. However, assuming we don't re-org through a finalized slot, it will never be
// *higher*.
//
// We provide this potentially-inaccurate-but-safe information to avoid onerous
// database reads during times of deep reorgs.
Ok(old_state
.finalized_checkpoint()
.epoch
.start_slot(E::slots_per_epoch()))
}
fn observe_head_block_delays<E: EthSpec, S: SlotClock>(
block_times_cache: &mut BlockTimesCache,
head_block: &ProtoBlock,
head_block_proposer_index: u64,
head_block_graffiti: String,
slot_clock: &S,
event_handler: Option<&ServerSentEventHandler<E>>,
log: &Logger,
) {
let block_time_set_as_head = timestamp_now();
let head_block_root = head_block.root;
let head_block_slot = head_block.slot;
let head_block_is_optimistic = head_block.execution_status.is_optimistic_or_invalid();
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
// Calculate the total delay between the start of the slot and when it was set as head.
let block_delay_total = get_slot_delay_ms(block_time_set_as_head, head_block_slot, slot_clock);
// Do not write to the cache for blocks older than 2 epochs, this helps reduce writes to
// the cache during sync.
if block_delay_total < slot_clock.slot_duration() * 64 {
block_times_cache.set_time_set_as_head(
head_block_root,
head_block_slot,
block_time_set_as_head,
);
}
// If a block comes in from over 4 slots ago, it is most likely a block from sync.
let block_from_sync = block_delay_total > slot_clock.slot_duration() * 4;
// Determine whether the block has been set as head too late for proper attestation
// production.
let late_head = block_delay_total >= slot_clock.unagg_attestation_production_delay();
// Do not store metrics if the block was > 4 slots old, this helps prevent noise during
// sync.
if !block_from_sync {
// Observe the total block delay. This is the delay between the time the slot started
// and when the block was set as head.
metrics::observe_duration(
&metrics::BEACON_BLOCK_HEAD_SLOT_START_DELAY_TIME,
block_delay_total,
);
// Observe the delay between when we imported the block and when we set the block as
// head.
let block_delays = block_times_cache.get_block_delays(
head_block_root,
slot_clock
.start_of(head_block_slot)
.unwrap_or_else(|| Duration::from_secs(0)),
);
metrics::observe_duration(
&metrics::BEACON_BLOCK_OBSERVED_SLOT_START_DELAY_TIME,
block_delays
.observed
.unwrap_or_else(|| Duration::from_secs(0)),
);
metrics::observe_duration(
&metrics::BEACON_BLOCK_HEAD_IMPORTED_DELAY_TIME,
block_delays
.set_as_head
.unwrap_or_else(|| Duration::from_secs(0)),
);
// If the block was enshrined as head too late for attestations to be created for it,
// log a debug warning and increment a metric.
if late_head {
metrics::inc_counter(&metrics::BEACON_BLOCK_HEAD_SLOT_START_DELAY_EXCEEDED_TOTAL);
debug!(
log,
"Delayed head block";
"block_root" => ?head_block_root,
"proposer_index" => head_block_proposer_index,
"slot" => head_block_slot,
"block_delay" => ?block_delay_total,
"observed_delay" => ?block_delays.observed,
"imported_delay" => ?block_delays.imported,
"set_as_head_delay" => ?block_delays.set_as_head,
);
}
}
if let Some(event_handler) = event_handler {
if !block_from_sync && late_head && event_handler.has_late_head_subscribers() {
let peer_info = block_times_cache.get_peer_info(head_block_root);
let block_delays = block_times_cache.get_block_delays(
head_block_root,
slot_clock
.start_of(head_block_slot)
.unwrap_or_else(|| Duration::from_secs(0)),
);
event_handler.register(EventKind::LateHead(SseLateHead {
slot: head_block_slot,
block: head_block_root,
peer_id: peer_info.id,
peer_client: peer_info.client,
proposer_index: head_block_proposer_index,
proposer_graffiti: head_block_graffiti,
block_delay: block_delay_total,
observed_delay: block_delays.observed,
imported_delay: block_delays.imported,
set_as_head_delay: block_delays.set_as_head,
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
execution_optimistic: head_block_is_optimistic,
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
}));
}
}
}