I have a three node HDFS cluster with a name-node. There is absolutely no IO going on this cluster or any jobs running on it and I just use it for testing the Disco HDFS integration. I noticed that two of the three data-nodes are using 100% CPU. They have been running for a long time (2 months) but with no jobs running on them and almost no usage:
$ hadoop version Hadoop 2.3.0 Subversion http://svn.apache.org/repos/asf/hadoop/common -r 1567123 Compiled by jenkins on 2014-02-11T13:40Z Compiled with protoc 2.5.0 >From source with checksum dfe46336fbc6a044bc124392ec06b85 Is this a known bug?