When having trouble with your Big Data cluster, it could be interesting to automatically generate some monitoring for your namenode.
When you have some support from the editor, they give you some tricks that help to investigate what is the problem with your namenode.
A little perl script could help you when you have some timeout during hdfs simple task.
You can find this Perl script here monitor-nn.pl
The usage is very simple, you can call this script every 5 minutes and set a timeout above which the script will dump the jmx thread in the .out log file.
So you can have the status of each thread, analyze them or give it to the support team for having help to analyze.
$ ./monitor-nn.pl 15 /run/hadoop/namenode.pid
I recommend you to use it for helping to know better the namenode activity.