We are using CDH 4.4 and the issue is, after every couple hours, both Jobtracker go into standby mode.
Even when I start the MR1 service, the JTs never go into Active mode and I use commandlne to put one of them in active mode.
After a few hours, without any errors in the logs, the JT goes back to standby mode. Stops listening on port 8021.
I tried increasing the maxclientConnection of zookeeper to 200, still the same.
How do I resolve this?