【kafka】Kafka leader -1

1.概述

转载：https://blog.csdn.net/u013933879/article/details/90055546

参考：Kafka无法消费！？究竟是bug的“沦陷”还是配置的“扭曲”？

2. 问题现象

在只有一个副本时，出现了broker节点丢失，因此导致了leader -1的情况

3.解决方案

直接修改zookeeper中该分区对应的state

set /brokers/topics/test_add_replication/partitions/1/state {"controller_epoch":140,"leader":1002,"version":1,"leader_epoch":12,"isr":[1002]}

稍等片刻，重新查看topic状态，即可发现leader已变更为1002

4.题外话

查看toipic状态命令

./kafka-topics.sh --describe --zookeeper localhost:2181 --topic test_add_rep
lication

查看kafka集群brokers命令

zookeeper/bin/zkCli.sh
ls /brokers/ids

修改topic的副本数量
PS：非常不建议只有1个副本

创建一个json文件，内容如下：

{
  "version": 1,
  "partitions": [
    {
      "topic": "test_add_replication",
      "partition": 0,
      "replicas": [
        1002,
        1004,
        1001
      ]
    },
    {
      "topic": "test_add_replication",
      "partition": 1,
      "replicas": [
        1002,
        1004,
        1001
      ]
    },
    {
      "topic": "test_add_replication",
      "partition": 2,
      "replicas": [
        1002,
        1004,
        1001
      ]
    },
    {
      "topic": "test_add_replication",
      "partition": 3,
      "replicas": [
        1002,
        1004,
        1001
      ]
    },
    {
      "topic": "test_add_replication",
      "partition": 4,
      "replicas": [
        1002,
        1004,
        1001
      ]
    }
  ]
}

执行修改命令

./kafka-reassign-partitions.sh --zookeeper localhost:2181 --reassignment-jso
n-file /tmp/rep.json --execute

修改完成后稍等片刻即可生效

5.威胁案例

【kafka】flink kafka NullPointerException Timeout position partition could be determined

6.正解

【Kafka】Kafka Leader:none ISR 为空 消费超时