Hadoop HA集群 NameNode 无法自动故障转移（切换active）

冷不防 2023-06-27 05:46 15阅读 0赞

**在学习 HA 自动化配置，按照hadoop官网:[https://hadoop.apache.org/docs/stable/hadoop-project-dist/hadoop-hdfs/HDFSHighAvailabilityWithQJM.html][https_hadoop.apache.org_docs_stable_hadoop-project-dist_hadoop-hdfs_HDFSHighAvailabilityWithQJM.html] 配置，最后所有的节点都启动正常。**

**用 kill -9 进程号 杀死了当前处于active状态的NameNode后，其他的 Standby 状态的NameNode 并没有自动切换为 Active状态，而且重启杀死的 NameNode 后，可能出现所有NameNode节点都变成了 Standby状态的情况，一个Active状态的都没有。**

**查看日志：$\{HADOOP\_HOME\}/logs/hadoop-root-zkfc-hadoop2.log ，发现报错了：**

2020-01-03 19:21:13,636 WARN org.apache.hadoop.ha.FailoverController: Unable to gracefully make NameNode at hadoop3/192.168.233.13:8020 standby (unable to connect)
    java.net.ConnectException: Call From hadoop2/192.168.233.12 to hadoop3:8020 failed on connection exception: java.net.ConnectException: 拒绝连接; For more details see:  http://wiki.apache.org/hadoop/ConnectionRefused
    	at sun.reflect.GeneratedConstructorAccessor8.newInstance(Unknown Source)
    	at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
    	at java.lang.reflect.Constructor.newInstance(Constructor.java:423)
    	at org.apache.hadoop.net.NetUtils.wrapWithMessage(NetUtils.java:792)
    .........................

**这是因为没有 fuster 程序，导致无法进行 fence，根据官网上的配置，是在 hdfs-site.xml 中配置过相关配置：**

<property>
        <name>dfs.ha.fencing.methods</name>
        <value>sshfence</value>
    </property>

**解决方法：安装包含fuster程序的软件包Psmisc（每个机器上都要安装）：**

yum -y install psmisc

**然后重启整个HA集群，进行测试，问题解决**

[https_hadoop.apache.org_docs_stable_hadoop-project-dist_hadoop-hdfs_HDFSHighAvailabilityWithQJM.html]: https://hadoop.apache.org/docs/stable/hadoop-project-dist/hadoop-hdfs/HDFSHighAvailabilityWithQJM.html