我在2节点集群上安装了CDH3U5。一切似乎运行良好,如所有服务,Web UI,MR作业,HDFS shell命令。然而,有趣的是,当我启动datanode服务时,它给了我一个OK消息,即datanode正在运行,因为进程说X.但是当我运行JPS时,我看不到该进程的标签“Datanode”。所以输出看起来像 -
17153 TaskTracker
18908 Jps
16267
进程ID-16267是Datanode进程。所有其他检查站都已通过。所以这看起来很怪异。同样的事情发生在集群中的另一个节点上。对此行为的任何了解以及是否需要修复都会有所帮助。
答案 0 :(得分:0)
can you check the following and reply? - web interface for namenode and what does it show there for livenode - logfiles for datanode to see if any exception - if datanode is pingable/ssh from namenode and viceversa If all the above look ok I'm not sure what the problem is but to fix you can - stop all hadoop deamons - delete temp directory pointed in conf/core-site.xml for both NN and DN - format namenode - start deamon