kafka分区迁移失败任务的处理

前一个文章"单副本kafka迁移分区"介绍了，怎样对分区进行迁移，然而在迁移最后一台broker的过程中，controler挂掉了，导致迁移任务一直没有进度，这里记录一下处理的过程

如果有一个迁移任务在执行时，重复执行，会有下面的提示，./bin/kafka-reassign-partitions.sh 是不能进行取消或者更改迁移任务状态的。

kafka.common.AdminCommandFailedException: Partition reassignment currently in progress for Map

 

处理思路：

1.登陆zk，把当前的任务删除

2.重新执行任务

 

具体操作：

2.1.登陆zk，删除任务

[zk: localhost:8181(CONNECTED) 2] ls /
[cluster, controller_epoch, controller, brokers, zookeeper, kafka-manager, admin, isr_change_notification, consumers, burrow, latest_producer_id_block, config]
[zk: localhost:8181(CONNECTED) 3] ls /admin/reassign_partitions
[]

#查看当前的任务
[zk: localhost:8181(CONNECTED) 5] get /admin/reassign_partitions 
{"version":1,"partitions":[{"topic":"yuanlog_json","partition":589,"replicas":[135]},{"topic":"………………
cZxid = 0x403a9ff6a
ctime = Thu Jul 25 10:50:12 CST 2019
mZxid = 0x403c68a74
mtime = Fri Jul 26 05:47:36 CST 2019
pZxid = 0x403a9ff6a
cversion = 0
dataVersion = 136189
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 9348
numChildren = 0

#删除任务
[zk: localhost:8181(CONNECTED) 8] rmr /admin/reassign_partitions

[work@kafka-node-134 kafka_2.11-0.10.0.0]$ ./bin/kafka-reassign-partitions.sh --zookeeper 10.69.59.186:8181 --reassignment-json-file ./tmp/136_exec0801.json --execute
Current partition replica assignment

2.2 重新执行迁移任务

[work@kafka-node-134 kafka_2.11-0.10.0.0]$ ./bin/kafka-reassign-partitions.sh --zookeeper 10.10.10.1:8181 --reassignment-json-file ./tmp/136_exec0801.json --execute
Current partition replica assignment

………………

Successfully started reassignment of partitions {"version":1,"partitions":[{"topic":"XX","partition":237,"replicas":[157]}]}

2.3 查看任务状态

[work@kafka-node-134 kafka_2.11-0.10.0.0]$ ./bin/kafka-reassign-partitions.sh --zookeeper 10.10.10.1:8181 --reassignment-json-file ./tmp/136_exec0801.json --verify