class kafka.common.NotLeaderForPartitionException(kafka.server.ReplicaFetcherThread)异常处理

[2017-06-15 19:33:14,716] ERROR [ReplicaFetcherThread-0-0], Error for partition [3_7_0_20170615,6] to broker 0:class kafka.common.NotLeaderForPartitionException (kafka.server.ReplicaFetcherThread)

[2017-06-15 19:33:14,718] ERROR [ReplicaFetcherThread-0-0], Error for partition [3_7_0_20170615,6] to broker 0:class kafka.common.NotLeaderForPartitionException(kafka.server.ReplicaFetcherThread)

执行

利用

tail -n300 kafkaServer.out | grepNotLeaderForPartitionException | awk -F"partition " '{print $2}' |awk -F"]" '{print $1}' | awk -F"[" '{print $2}'| sort |uniq | sort

制作json文件（node005 /opt/isu/kafka/bin下有）

{ 

 "partitions":

  [ 

   {"topic":"2_0_2_20170618","partition":21},

   {"topic":"3_0_20170618","partition":19},

   {"topic":"2_0_1_hdfs_8089_kafka_2_20170618","partition":44},

   {"topic":"3_7_0_hdfs_8089_kafka_20170618","partition":13},

   {"topic":"sdk_HDFS_168_20170618","partition":44}

  ] 

}

只均衡个别topic的个别partition

kafka-preferred-replica-election.sh --zookeeper node001:2181,node002:2181,node003:2181/kafka --path-to-json-file  ./partition_elec.json

如果有太多topic报错，建议直接对所有topic进行：

kafka-preferred-replica-election.sh --zookeeper node001:2181,node002:2181,node003:2181/kafka