I'm using CM v5.4.4 and I noticed this morning that the cluster had slowed down. The queries are taking longer to execute and even doing a simple
list of the data (hdfs dfs -ls /) is taking more than 20seconds (it was quicker before).
Doing a top on the DN and the NN's, I noticed that the cm-agent was taking about 150% of the CPU. Also, in the YARN logs, I noticed a lot of the
2015-10-05 15:33:37,312 INFO org.apache.hadoop.yarn.server.resourcemanager.scheduler.fair.FairScheduler: Null container completed...