Kafka Dependency on ZooKeeper Latency Cascade
criticalApache Kafka (pre-KRaft) stores critical metadata in ZooKeeper including broker registrations, topic configurations, and controller election state. High ZooKeeper latency or unavailability directly impacts Kafka broker operations and can cause broker flapping or topic unavailability.
When monitoring both Kafka and ZooKeeper, correlate Kafka broker registration changes or controller election events with ZooKeeper latency spikes and session expiration events. Alert when ZooKeeper latency exceeds 50ms while Kafka shows increased controller elections or broker state changes.
Prioritize ZooKeeper health when running Kafka - ensure dedicated disk for ZooKeeper transaction logs, monitor ZooKeeper ensemble quorum status, and set appropriate session timeouts. Consider migrating to Kafka KRaft mode to eliminate ZooKeeper dependency. If staying with ZooKeeper, ensure it runs on dedicated nodes separate from Kafka brokers.