Hello, I have a job with no reducers and 20 mappers, out of which 14 get scheduled at a time. But strangely even if mappers reach a SUCEEDED status the container running them is not killed. Due to which the next set of mappers never get scheduled. I admit that there may be threads running in the jvm which can be stopped but shouldnt a SIGTERM be issued to container automatically?
I am on hadoop 2.2.0 and cascading 2.5.5 on debian wheezy*. Anyone else faced this? Regards