选举前：{ \"controller_epoch\
选举后：{ \"controller_epoch\
1、根据相应的策略选出leader 2、将选举的结果写入zk中，对应节点为brokers/topics/{topic}/partitions/{partition}/state 3、更新controllerContext中的partitionLeadershipInfo信息 4、给每个还存活的副本发送leaderAndIsr请求 5、给集群中每个broker发送updateMetadata请求
controller
watch
leaderElectionClient
LeaderElection脚本
topic_1-0
state
LeaderElectionCommand
LeaderElection脚本：{ \"partitions\": [ { \"topic\": \"topic_1\
topic_1
kafka cluster
broker1
admin
校验数据
写入zk节点数据：{ \"partitions\": [ { \"topic\": \"topic_1\
leaderAndIsrRequest
0
调用electLeaders
partitions
KafkaController
写入
调用分区状态机处理分区状态
topics
{ \"version\
zk
brokers
broker3
broker2
preferred_replica_election
write
