Ensure custody backfill sync couples all responses before importing (#8339)

Custody backfill sync has a bug when we request columns from more than one peer per batch. The fix here ensures we wait for all requests to be completed before performing verification and importing the responses.

I've also added an endpoint `lighthouse/custody/backfill` that resets a nodes earliest available data column to the current epoch so that custody backfill can be triggered. This endpoint is needed to rescue any nodes that may have missing columns due to the custody backfill sync bug without requiring a full re-sync.


  


Co-Authored-By: Eitan Seri- Levi <eserilev@gmail.com>

Co-Authored-By: Eitan Seri-Levi <eserilev@ucsc.edu>

Co-Authored-By: Jimmy Chen <jchen.tc@gmail.com>

Co-Authored-By: Michael Sproul <michaelsproul@users.noreply.github.com>
This commit is contained in:
Eitan Seri-Levi
2025-11-03 00:06:06 -08:00
committed by GitHub
parent 4908687e7d
commit 5d0f8a083a
11 changed files with 230 additions and 24 deletions

View File

@@ -4604,6 +4604,37 @@ pub fn serve<T: BeaconChainTypes>(
},
);
// POST lighthouse/custody/backfill
let post_lighthouse_custody_backfill = warp::path("lighthouse")
.and(warp::path("custody"))
.and(warp::path("backfill"))
.and(warp::path::end())
.and(task_spawner_filter.clone())
.and(chain_filter.clone())
.then(
|task_spawner: TaskSpawner<T::EthSpec>, chain: Arc<BeaconChain<T>>| {
task_spawner.blocking_json_task(Priority::P1, move || {
// Calling this endpoint will trigger custody backfill once `effective_epoch``
// is finalized.
let effective_epoch = chain
.canonical_head
.cached_head()
.head_slot()
.epoch(T::EthSpec::slots_per_epoch())
+ 1;
let custody_context = chain.data_availability_checker.custody_context();
// Reset validator custody requirements to `effective_epoch` with the latest
// cgc requiremnets.
custody_context.reset_validator_custody_requirements(effective_epoch);
// Update `DataColumnCustodyInfo` to reflect the custody change.
chain.update_data_column_custody_info(Some(
effective_epoch.start_slot(T::EthSpec::slots_per_epoch()),
));
Ok(())
})
},
);
// GET lighthouse/analysis/block_rewards
let get_lighthouse_block_rewards = warp::path("lighthouse")
.and(warp::path("analysis"))
@@ -4963,6 +4994,7 @@ pub fn serve<T: BeaconChainTypes>(
.uor(post_lighthouse_compaction)
.uor(post_lighthouse_add_peer)
.uor(post_lighthouse_remove_peer)
.uor(post_lighthouse_custody_backfill)
.recover(warp_utils::reject::handle_rejection),
),
)

View File

@@ -1,6 +1,7 @@
use crate::{Config, Context};
use beacon_chain::{
BeaconChain, BeaconChainTypes,
custody_context::NodeCustodyType,
test_utils::{BeaconChainHarness, BoxedMutator, Builder, EphemeralHarnessType},
};
use beacon_processor::{
@@ -67,6 +68,20 @@ impl<E: EthSpec> InteractiveTester<E> {
None,
Config::default(),
true,
NodeCustodyType::Fullnode,
)
.await
}
pub async fn new_supernode(spec: Option<ChainSpec>, validator_count: usize) -> Self {
Self::new_with_initializer_and_mutator(
spec,
validator_count,
None,
None,
Config::default(),
true,
NodeCustodyType::Supernode,
)
.await
}
@@ -78,6 +93,7 @@ impl<E: EthSpec> InteractiveTester<E> {
mutator: Option<Mutator<E>>,
config: Config,
use_mock_builder: bool,
node_custody_type: NodeCustodyType,
) -> Self {
let mut harness_builder = BeaconChainHarness::builder(E::default())
.spec_or_default(spec.map(Arc::new))
@@ -93,6 +109,8 @@ impl<E: EthSpec> InteractiveTester<E> {
.fresh_ephemeral_store()
};
harness_builder = harness_builder.node_custody_type(node_custody_type);
// Add a mutator for the beacon chain builder which will be called in
// `HarnessBuilder::build`.
if let Some(mutator) = mutator {

View File

@@ -1,3 +1,4 @@
use beacon_chain::custody_context::NodeCustodyType;
use beacon_chain::test_utils::test_spec;
use beacon_chain::{
GossipVerifiedBlock, IntoGossipVerifiedBlock, WhenSlotSkipped,
@@ -1956,6 +1957,7 @@ pub async fn duplicate_block_status_code() {
..Config::default()
},
true,
NodeCustodyType::Fullnode,
)
.await;

View File

@@ -1,4 +1,5 @@
//! Tests for API behaviour across fork boundaries.
use beacon_chain::custody_context::NodeCustodyType;
use beacon_chain::{
StateSkipConfig,
test_utils::{DEFAULT_ETH1_BLOCK_HASH, HARNESS_GENESIS_TIME, RelativeSyncCommittee},
@@ -426,6 +427,7 @@ async fn bls_to_execution_changes_update_all_around_capella_fork() {
None,
Default::default(),
true,
NodeCustodyType::Fullnode,
)
.await;
let harness = &tester.harness;

View File

@@ -1,4 +1,5 @@
//! Generic tests that make use of the (newer) `InteractiveApiTester`
use beacon_chain::custody_context::NodeCustodyType;
use beacon_chain::{
ChainConfig,
chain_config::{DisallowedReOrgOffsets, ReOrgThreshold},
@@ -76,6 +77,7 @@ async fn state_by_root_pruned_from_fork_choice() {
None,
Default::default(),
false,
NodeCustodyType::Fullnode,
)
.await;
@@ -433,6 +435,7 @@ pub async fn proposer_boost_re_org_test(
})),
Default::default(),
false,
NodeCustodyType::Fullnode,
)
.await;
let harness = &tester.harness;
@@ -1049,6 +1052,68 @@ async fn proposer_duties_with_gossip_tolerance() {
);
}
// Test that a request to `lighthouse/custody/backfill` succeeds by verifying that `CustodyContext` and `DataColumnCustodyInfo`
// have been updated with the correct values.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn lighthouse_restart_custody_backfill() {
let spec = test_spec::<E>();
// Skip pre-Fulu.
if !spec.is_fulu_scheduled() {
return;
}
let validator_count = 24;
let tester = InteractiveTester::<E>::new_supernode(Some(spec), validator_count).await;
let harness = &tester.harness;
let spec = &harness.spec;
let client = &tester.client;
let min_cgc = spec.custody_requirement;
let max_cgc = spec.number_of_custody_groups;
let num_blocks = 2 * E::slots_per_epoch();
let custody_context = harness.chain.data_availability_checker.custody_context();
harness.advance_slot();
harness
.extend_chain_with_sync(
num_blocks as usize,
BlockStrategy::OnCanonicalHead,
AttestationStrategy::AllValidators,
SyncCommitteeStrategy::NoValidators,
LightClientStrategy::Disabled,
)
.await;
let cgc_at_head = custody_context.custody_group_count_at_head(spec);
let earliest_data_column_epoch = harness.chain.earliest_custodied_data_column_epoch();
assert_eq!(cgc_at_head, max_cgc);
assert_eq!(earliest_data_column_epoch, None);
custody_context
.update_and_backfill_custody_count_at_epoch(harness.chain.epoch().unwrap(), cgc_at_head);
client.post_lighthouse_custody_backfill().await.unwrap();
let cgc_at_head = custody_context.custody_group_count_at_head(spec);
let cgc_at_previous_epoch =
custody_context.custody_group_count_at_epoch(harness.chain.epoch().unwrap() - 1, spec);
let earliest_data_column_epoch = harness.chain.earliest_custodied_data_column_epoch();
// `DataColumnCustodyInfo` should have been updated to the head epoch
assert_eq!(
earliest_data_column_epoch,
Some(harness.chain.epoch().unwrap() + 1)
);
// Cgc requirements should have stayed the same at head
assert_eq!(cgc_at_head, max_cgc);
// Cgc requirements at the previous epoch should be `min_cgc`
// This allows for custody backfill to re-fetch columns for this epoch.
assert_eq!(cgc_at_previous_epoch, min_cgc);
}
// Test that a request for next epoch proposer duties suceeds when the current slot clock is within
// gossip clock disparity (500ms) of the new epoch.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]