When looking for peers, skip peers with error instead of aborting the whole function (#15815)

* `findPeersWithSubnets`: If the `filter` function returns an error for a given peer, log an error and skip the peer instead of aborting the whole function.

* `computeIndicesByRootByPeer`: If the loop returns an error for a given peer, log an error and skip the peer instead of aborting the whole function.

* Add changelog.

---------

Co-authored-by: james-prysm <90280386+james-prysm@users.noreply.github.com>

beacon-chain/p2p/subnets.go

@@ -25,6 +25,7 @@ import (
 	"github.com/holiman/uint256"
 	"github.com/pkg/errors"
 	"github.com/prysmaticlabs/go-bitfield"
+	"github.com/sirupsen/logrus"
 )
 
 var (
@@ -223,8 +224,14 @@ func (s *Service) findPeersWithSubnets(
 		// Skip nodes that are not subscribed to any of the defective subnets.
 		nodeSubnets, err := filter(node)
 		if err != nil {
-			return nil, errors.Wrap(err, "filter node")
+			log.WithError(err).WithFields(logrus.Fields{
+				"nodeID":      node.ID(),
+				"topicFormat": topicFormat,
+			}).Debug("Could not get needed subnets from peer")
+
+			continue
 		}
+
 		if len(nodeSubnets) == 0 {
 			continue
 		}

beacon-chain/sync/data_column_sidecars.go

@@ -1022,17 +1022,20 @@ func computeIndicesByRootByPeer(
 	peersByIndex := make(map[uint64]map[goPeer.ID]bool)
 	headSlotByPeer := make(map[goPeer.ID]primitives.Slot)
 	for peer := range peers {
+		log := log.WithField("peerID", peer)
+
 		// Computes the custody columns for each peer
 		nodeID, err := prysmP2P.ConvertPeerIDToNodeID(peer)
 		if err != nil {
-			return nil, errors.Wrapf(err, "convert peer ID to node ID for peer %s", peer)
+			log.WithError(err).Debug("Failed to convert peer ID to node ID")
+			continue
 		}
 
 		custodyGroupCount := p2p.CustodyGroupCountFromPeer(peer)
-
 		dasInfo, _, err := peerdas.Info(nodeID, custodyGroupCount)
 		if err != nil {
-			return nil, errors.Wrapf(err, "peerdas info for peer %s", peer)
+			log.WithError(err).Debug("Failed to get peer DAS info")
+			continue
 		}
 
 		for column := range dasInfo.CustodyColumns {
@@ -1045,11 +1048,13 @@ func computeIndicesByRootByPeer(
 		// Compute the head slot for each peer
 		peerChainState, err := p2p.Peers().ChainState(peer)
 		if err != nil {
-			return nil, errors.Wrapf(err, "get chain state for peer %s", peer)
+			log.WithError(err).Debug("Failed to get peer chain state")
+			continue
 		}
 
 		if peerChainState == nil {
-			return nil, errors.Errorf("chain state is nil for peer %s", peer)
+			log.Debug("Peer chain state is nil")
+			continue
 		}
 
 		// Our view of the head slot of a peer is not updated in real time.

changelog/manu-skip-bad-peers.md

@@ -0,0 +1,3 @@
+### Fixed 
+- `findPeersWithSubnets`: If the filter function returns an error for a given peer, log an error and skip the peer instead of aborting the whole function.
+- `computeIndicesByRootByPeer`: If the loop returns an error for a given peer, log an error and skip the peer instead of aborting the whole function.