ZhenChun Pan created KAFKA-13564: ------------------------------------ Summary: Kafka keep print NOT_LEADER_OR_FOLLOWER in log file after one broker dropped, and the producer can not work. Key: KAFKA-13564 URL: https://issues.apache.org/jira/browse/KAFKA-13564 Project: Kafka Issue Type: Bug Affects Versions: 2.7.0 Reporter: ZhenChun Pan
The machine of broker0 dropped, and some partition change the leader to broker1. We can found message like below in state-change.log: [2021-12-11 15:34:14,868] TRACE [Broker id=0] Cached leader info UpdateMetadataPartitionState(topicName='ceae-1002-flink-characteristic-instance-data', partitionIndex=0, controllerEpoch=3, leader=1, leaderEpoch=8, isr=[1], zkVersion*****#*#*****, offlineReplicas=[]) for partition ceae-1002-flink-characteristic-instance-data-0 in response to UpdateMetadata request sent by controller 2 epoch 3 with correlation id 0 (state.change.logger) But we found server.log keep print logs like below: [2021-12-11 15:34:30,272] INFO [ReplicaFetcher replicaId=0, leaderId=1, fetcherId=6] Retrying leaderEpoch request for partition ceae-1002-flink-characteristic-instance-data-0 as the leader reported an error: NOT_LEADER_OR_FOLLOWER (kafka.server.ReplicaFetcherThread) And the producer also can not work and print messages below: [2021-12-11 16:00:00,703] INFO [ReplicaFetcher replicaId=0, leaderId=1, fetcherId=4] Retrying leaderEpoch request for partition ceae-1002-flink-characteristic-instance-data-0 as the leader reported an error: NOT_LEADER_OR_FOLLOWER (kafka.server.ReplicaFetcherThread) We resume broker0, but did not work. So we restart all brokers of kafka cluster, and fix the trouble. -- This message was sent by Atlassian Jira (v8.20.1#820001)