Page 1 of 1

PX job memory leak...

Posted: Wed Oct 04, 2006 2:12 pm
by pneumalin
Friends,
Does anyone know how to dectect the Memory Leak caused PX jobs and clean it up in runtime...
Why I am asking for this?
Since we had a PX job using Enterprise DB2 stage to perform load to Remote DB2 Server, and this job constantly Hangs to timeout in 2 hours for no reason. But it load through once it is restarted.. Our DBA was monitoring the DB2 activity at that time and found out there is no data written to the Named-Pipe for 2 hours where DB2 Loader is trying to read.
Therefore we suspect the PX Engines lost the communnication link at that moment to cause the Hang. Furthermore, the restart works ensure us there must be a resource issue in DataStage Engine runtime and it get cleared once the Job is restarted..
Just try to find a way to debug the problem, any advise is welcome!
Thanks.

Posted: Wed Oct 04, 2006 3:14 pm
by ray.wurlod
Have you specified "all rows in single transaction"?

Is it really a memory leak or just consuming a lot of memory that it will relinquish once it's done? A memory leak is a bug, using a lot of memory is not.

Posted: Wed Oct 04, 2006 3:33 pm
by pneumalin
Ray,
Is "All row in a single transaction" a DB2 setup option for DBA? Or we can set it from Enterprise DB2/UDB Stage? I don't see this option in that Stage though? I am running 7.5.0.
Please advise.
Thanks!