gnosis/lighthouse
1
0
Fork 0
mirror of https://github.com/sigp/lighthouse.git synced 2026-03-19 21:04:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

Backfill peer attribution (#7762)

Browse Source 
Partly addresses https://github.com/sigp/lighthouse/issues/7744


  Implement similar peer sync attribution like in #7733 for backfill sync.
This commit is contained in:
Pawan Dhananjay
2025-08-11 19:11:56 -07:00
committed by GitHub
parent 122f16776f
commit 80ba0b169b
7 changed files with 620 additions and 144 deletions
Download Patch File Download Diff File Expand all files Collapse all files

77
beacon_node/network/src/sync/range_sync/chain.rs
View File

@@ -817,32 +817,44 @@ impl<T: BeaconChainTypes> SyncingChain<T> {
) -> ProcessingResult {
let batch_state = self.visualize_batch_state();
if let Some(batch) = self.batches.get_mut(&batch_id) {
if let RpcResponseError::BlockComponentCouplingError(CouplingError {
column_and_peer,
msg,
}) = &err
{
debug!(?batch_id, msg, "Block components coupling error");
// Note: we don't fail the batch here because a `CouplingError` is
// recoverable by requesting from other honest peers.
if let Some((column_and_peer, action)) = column_and_peer {
let mut failed_columns = HashSet::new();
let mut failed_peers = HashSet::new();
for (column, peer) in column_and_peer {
failed_columns.insert(*column);
failed_peers.insert(*peer);
if let RpcResponseError::BlockComponentCouplingError(coupling_error) = &err {
match coupling_error {
CouplingError::DataColumnPeerFailure {
error,
faulty_peers,
action,
exceeded_retries,
} => {
debug!(?batch_id, error, "Block components coupling error");
// Note: we don't fail the batch here because a `CouplingError` is
// recoverable by requesting from other honest peers.
let mut failed_columns = HashSet::new();
let mut failed_peers = HashSet::new();
for (column, peer) in faulty_peers {
failed_columns.insert(*column);
failed_peers.insert(*peer);
}
for peer in failed_peers.iter() {
network.report_peer(*peer, *action, "failed to return columns");
}
// Retry the failed columns if the column requests haven't exceeded the
// max retries. Otherwise, remove treat it as a failed batch below.
if !*exceeded_retries {
return self.retry_partial_batch(
network,
batch_id,
request_id,
failed_columns,
failed_peers,
);
}
}
for peer in failed_peers.iter() {
network.report_peer(*peer, *action, "failed to return columns");
CouplingError::BlobPeerFailure(msg) => {
tracing::debug!(?batch_id, msg, "Blob peer failure");
}
CouplingError::InternalError(msg) => {
tracing::error!(?batch_id, msg, "Block components coupling internal error");
}
return self.retry_partial_batch(
network,
batch_id,
request_id,
failed_columns,
failed_peers,
);
}
}
// A batch could be retried without the peer failing the request (disconnecting/
@@ -900,14 +912,11 @@ impl<T: BeaconChainTypes> SyncingChain<T> {
let (request, batch_type) = batch.to_blocks_by_range_request();
let failed_peers = batch.failed_peers();
// TODO(das): we should request only from peers that are part of this SyncingChain.
// However, then we hit the NoPeer error frequently which causes the batch to fail and
// the SyncingChain to be dropped. We need to handle this case more gracefully.
let synced_peers = network
.network_globals()
.peers
.read()
.synced_peers_for_epoch(batch_id, &self.peers)
.synced_peers_for_epoch(batch_id, Some(&self.peers))
.cloned()
.collect::<HashSet<_>>();
@@ -984,7 +993,7 @@ impl<T: BeaconChainTypes> SyncingChain<T> {
.network_globals()
.peers
.read()
.synced_peers()
.synced_peers_for_epoch(batch_id, Some(&self.peers))
.cloned()
.collect::<HashSet<_>>();
@@ -1084,11 +1093,13 @@ impl<T: BeaconChainTypes> SyncingChain<T> {
.sampling_subnets()
.iter()
.all(|subnet_id| {
let peer_count = network
.network_globals()
let peer_db = network.network_globals().peers.read();
let peer_count = self
.peers
.read()
.good_range_sync_custody_subnet_peer(*subnet_id, &self.peers)
.iter()
.filter(|peer| {
peer_db.is_good_range_sync_custody_subnet_peer(*subnet_id, peer)
})
.count();
peer_count > 0
});