Details
Description
We are running CDH 5.14.2 on a mix of 12.04 and 16.04 machines. We are seeing some containers continue to run even after the Map reduce application has completed successfully. From the logs of the NodeManagers you can see that it is trying to kill the container but it does not get the process exit status. We did strace on the NodeManager and found it was trying to to issue commands like this:
kill -0 – -1234
However on 12.04 hosts the kill command does not support the two dashesh. As a work around we did a kill script that supports this new syntax.