I have two type of jobs that require the cluster to have different number of tasks per node. After doing the research, I end up with this page:
https://issues.apache.org/jira/browse/HADOOP-5170
I am wondering if anybody else knows ways to walk around. I don't want to restart the cluster every time.
Or if anybody can tell me how bad the performance will be if I set a bad number for it, let's say, 1 or 999999 tasks per node.
Thanks,