Skip to content

Kafka Troubleshooting

Overview

1. Metadata / connection failures

Symptoms

Connection to node -1 could not be established or unreachable brokers from Metadata.

Keywords

advertised.listeners / NAT / container networking

Steps

  1. Verify bootstrap addresses are reachable.
  2. Check advertised.listeners on each broker.
  3. Test every host:port returned in Metadata.
  4. Align security protocol with listener.security.protocol.map.

Reference

2. Consumer lag keeps growing

Steps

  1. Check consumer process health and rebalance storms.
  2. Match partition count to consumer parallelism.
  3. Review max.poll.interval.ms and processing time.
  4. Inspect broker disk/network and ISR shrink.

3. Under-Replicated Partitions (URP)

Steps

  1. kafka-topics.sh --describe for URP partitions.
  2. Check offline replica brokers (disk, GC, network).
  3. Tune replica.lag.time.max.ms for cross-DC RTT.
  4. Reassign partitions after replicas recover.

4. Disk full / retention issues

Steps

  1. kafka-log-dirs.sh --describe for large topics.
  2. Compare broker defaults vs topic-level retention overrides.
  3. Verify cleanup.policy for compact topics.
  4. Emergency: expand disk or adjust retention per change process.