From fe9dd255c7054b8bff32fbbcced68db34b75ff15 Mon Sep 17 00:00:00 2001
From: Preston Van Loon <pvanloon@offchainlabs.com>
Date: Thu, 25 Sep 2025 10:46:28 -0500
Subject: [PATCH] slasherkv: Set a 1 minute timeout on PruneAttestationOnEpoch
 operations (#15746)

* slasherkv: Set a 1 minute timeout on PruneAttestationOnEpoch operations to prevent very large bolt transactions.

* Fix CI

---------

Co-authored-by: Manu NALEPA <enalepa@offchainlabs.com>
---
 beacon-chain/db/slasherkv/pruning.go | 33 +++++++++++++++++++++++-----
 changelog/pvl-slasherkv-timeout.md   |  2 ++
 2 files changed, 30 insertions(+), 5 deletions(-)
 create mode 100644 changelog/pvl-slasherkv-timeout.md

diff --git a/beacon-chain/db/slasherkv/pruning.go b/beacon-chain/db/slasherkv/pruning.go
index 1334d375c8..8f4e2efbe5 100644
--- a/beacon-chain/db/slasherkv/pruning.go
+++ b/beacon-chain/db/slasherkv/pruning.go
@@ -4,17 +4,26 @@ import (
 	"bytes"
 	"context"
 	"encoding/binary"
+	"time"
 
 	"github.com/OffchainLabs/prysm/v6/consensus-types/primitives"
 	"github.com/OffchainLabs/prysm/v6/time/slots"
+	"github.com/pkg/errors"
 	bolt "go.etcd.io/bbolt"
 )
 
+var errTimeOut = errors.New("operation timed out")
+
 // PruneAttestationsAtEpoch deletes all attestations from the slasher DB with target epoch
 // less than or equal to the specified epoch.
 func (s *Store) PruneAttestationsAtEpoch(
-	_ context.Context, maxEpoch primitives.Epoch,
+	ctx context.Context, maxEpoch primitives.Epoch,
 ) (numPruned uint, err error) {
+	// In some cases, pruning may take a very long time and consume significant memory in the
+	// open Update transaction. Therefore, we impose a 1 minute timeout on this operation.
+	ctx, cancel := context.WithTimeout(ctx, 1*time.Minute)
+	defer cancel()
+
 	// We can prune everything less than the current epoch - history length.
 	encodedEndPruneEpoch := make([]byte, 8)
 	binary.BigEndian.PutUint64(encodedEndPruneEpoch, uint64(maxEpoch))
@@ -48,13 +57,18 @@ func (s *Store) PruneAttestationsAtEpoch(
 		return
 	}
 
-	if err = s.db.Update(func(tx *bolt.Tx) error {
+	err = s.db.Update(func(tx *bolt.Tx) error {
 		signingRootsBkt := tx.Bucket(attestationDataRootsBucket)
 		attRecordsBkt := tx.Bucket(attestationRecordsBucket)
 		c := signingRootsBkt.Cursor()
 
 		// We begin a pruning iteration starting from the first item in the bucket.
 		for k, v := c.First(); k != nil; k, v = c.Next() {
+			if ctx.Err() != nil {
+				// Exit the routine if the context has expired.
+				return errTimeOut
+			}
+
 			// We check the epoch from the current key in the database.
 			// If we have hit an epoch that is greater than the end epoch of the pruning process,
 			// we then completely exit the process as we are done.
@@ -67,18 +81,27 @@ func (s *Store) PruneAttestationsAtEpoch(
 			// so it is possible we have a few adjacent objects that have the same slot, such as
 			//  (target_epoch = 3 ++ _) => encode(attestation)
 			if err := signingRootsBkt.Delete(k); err != nil {
-				return err
+				return errors.Wrap(err, "delete attestation signing root")
 			}
 			if err := attRecordsBkt.Delete(v); err != nil {
-				return err
+				return errors.Wrap(err, "delete attestation record")
 			}
 			slasherAttestationsPrunedTotal.Inc()
 			numPruned++
 		}
 		return nil
-	}); err != nil {
+	})
+
+	if errors.Is(err, errTimeOut) {
+		log.Warning("Aborting pruning routine")
 		return
 	}
+
+	if err != nil {
+		log.WithError(err).Error("Failed to prune attestations")
+		return
+	}
+
 	return
 }
 
diff --git a/changelog/pvl-slasherkv-timeout.md b/changelog/pvl-slasherkv-timeout.md
new file mode 100644
index 0000000000..7f632f8f4f
--- /dev/null
+++ b/changelog/pvl-slasherkv-timeout.md
@@ -0,0 +1,2 @@
+### Added
+- Added a 1 minute timeout on PruneAttestationOnEpoch operations to prevent very large bolt transactions