启动yarn时，resourcemanager启动失败

问题描述

使用sbin/start-yarn.sh启动yarn后，再使用jps查看进程，发现只有Notemanager启动了，而resourcemanager没有启动。
代码如下所示：

[hadoop@hadoop-01 hadoop]$ sbin/start-yarn.sh
starting yarn daemons
starting resourcemanager, logging to /home/hadoop/app/hadoop-2.6.0-cdh5.10.0/logs/yarn-hadoop-resourcemanager-hadoop-01.out
hadoop-01: starting nodemanager, logging to /home/hadoop/app/hadoop-2.6.0-cdh5.10.0/logs/yarn-hadoop-nodemanager-hadoop-01.out
hadoop-02: starting nodemanager, logging to /home/hadoop/app/hadoop-2.6.0-cdh5.10.0/logs/yarn-hadoop-nodemanager-hadoop-02.out
hadoop-03: starting nodemanager, logging to /home/hadoop/app/hadoop-2.6.0-cdh5.10.0/logs/yarn-hadoop-nodemanager-hadoop-03.out
[hadoop@hadoop-01 hadoop]$ jps
15824 Jps
2241 QuorumPeerMain
13461 NameNode
14054 DFSZKFailoverController
13575 DataNode
13771 JournalNode
15691 NodeManager

---

原因分析：

先查看日志文件，启动resourcemanager的日志文件的位置如下图红框所示，将后缀.out改为.log就是日志文件。

#查看日志
cat /home/hadoop/app/hadoop-2.6.0-cdh5.10.0/logs/yarn-hadoop-resourcemanager-hadoop-01.log

在日志中发现错误原因是：yarn.resourcemanager.zk-address is not set，问题应该是在yarn的配置文件中没有找到yarn.resourcemanager.zk-address

解决方案：

打开yarn的配置文件yarn-site.xml，找到如下图所示的配置段：

将hadoop.zk.address改为yarn.resourcemanager.zk-address
随后重新启动yarn，发现resourcemanager进程已经存在。