i find the same situation in stackoverflow: http://stackoverflow.com/questions/35857130/kafka-broker-removed-from-zookeeper-while-leader-election-occurs-error but there are no explaination
2016-05-18 12:02 GMT+08:00 Fredo Lee <buaatianwa...@gmail.com>: > hi all > > we have 20 kafka brokers in on cluster. and this morning, we cannot find > one kafka broker when using "kafka-topic.sh" to get topic information. > > we find lots of error: "[2016-05-18 09:59:59,998] ERROR > [ReplicaFetcherThread-0-1005], Error for partition [0000000033,58] to > broker 1005:class kafka.common.NotLeaderForPartitionException > (kafka.server.ReplicaFetcherThread)" in server.log in the wrong status > broker. > > this broker is running always and the state-change.log is clear. > > finnaly, we restart this broker and it works. > > > who can explain this scene? what's the reason? >