Recently, we are facing "namenode's leasemanager from looping forever". It will bring namenode to a state nothing could be executed
other than rolling it.
It looks like is a known issue fixed in later versions as per the ticket https://issues.apache.org/jira/browse/HDFS-4882.
We are currently running cdh4.7.1.p0.47, the fix is available in CDH 5.4/apache 2.6.1. In this jira, there is a patch for this issue that has
been tested in later version. I wanted to check if anybody who faced this problem tried to use this patch and compile cdh4.7.1 code and
was the issue fixed?
This issue stopped happening after few hardware upgrades to name node. We thought of spending time in upgrading to cdh5 instead of patch to cdh4.