故障现象：

[root@AY1312280620257702e0Z ~]# jps

6822 NodeManager

6106 SecondaryNameNode

6561 ResourceManager

6005 NameNode

6955 JobHistoryServer

10964 Jps

Without datanode

[root@AY1312280620257702e0Z ~]# hadoop fs -mkdir input

[root@AY1312280620257702e0Z ~]# hadoop fs -put /etc/hadoop/conf/*.xml input

14/01/28 22:50:45 WARN hdfs.DFSClient: DataStreamer Exception

org.apache.hadoop.ipc.RemoteException(java.io.IOException): File /user/root/input/core-site.xml._COPYING_ could only be replicated to 0 nodes instead of minReplication (=1). There are 0 datanode(s) running and no node(s) are excluded in this operation.

问题定位：

检查日志/var/log/hadoop-hdfs/hadoop-hdfs-datanode-AY1312280620257702e0Z.log.发现有如下异常：

2014-01-28 20:52:16,377 FATAL org.apache.hadoop.hdfs.server.datanode.DataNode: Initialization failed for block pool Block pool BP-1814315425-10.144.162.11-1390913469801 (storage id DS-938709897-10.144.162.11-50010-1390910093924) service to localhost/127.0.0.1:8020

java.io.IOException: Incompatible clusterIDs in /var/lib/hadoop-hdfs/cache/hdfs/dfs/data: namenode clusterID = CID-398d1409-5590-4bed-a8e2-ee731cf3d508; datanode clusterID = CID-28de7f67-d405-42fb-9dab-dbc69eb8f0b2

根因：

因为多次执行了"sudo -u hdfs hdfs namenode -format" ，而没有手动清除cd /var/lib/hadoop-hdfs/cache/hdfs/dfs/data路径下的多余文件，当手动删除该路径下的文件，重新运行启动HDFS的命令，则HDFS DATANODE可以运行。

ITPub博客

[原创]安装CDH5 BETA的时候，无法启动DATANODE