Apache Kafka

Raft Commit Latency Spike Delays Metadata Propagation

warning
latencyUpdated Mar 2, 2026

In KRaft mode, high Raft commit latency delays metadata changes from propagating through the cluster, causing stale metadata and operational delays.

Technologies:
How to detect:

Monitor kafka.raft.commit_latency_avg or kafka.raft.commit_latency_max exceeding 1 second. Cross-reference with kafka.raft.append_records_rate to check write volume.

Recommended action:

1. Check controller disk I/O: Verify metadata log storage performance. 2. Review controller CPU: Ensure controllers have sufficient CPU capacity. 3. Check network latency: Verify inter-controller network performance. 4. Monitor quorum size: Larger quorums have higher commit latency. 5. Review metadata change rate: Excessive metadata changes can overwhelm controllers. 6. Optimize client operations: Reduce frequency of metadata-changing operations.