DataStage OSH system process PID Tracking

newsaur · Post by **newsaur** » Thu Jun 15, 2006 12:25 am

We are facing memory contention problems. With SVMON we have identified the exact system processes that are eating up gigabytes of memories (which is not expected in our development memory with very little data).

With the UNIX process ID (PID), is there any way to track exactly to which DataStage Job instance it belongs?

Thanks.

ArndW · Post by **ArndW** » Thu Jun 15, 2006 12:46 am

I'm not at a DataStage client right now so cannot check up, but I thought either setting APT_DUMP_SCORE or another debugging switch will display not only how many processes are used in running a job but also which PIDs they have. You can also traceback the PID to its parent and so on until you get to the UV process that originated it, and that will have the job name in its "ps -ef" display.

ray.wurlod · Post by **ray.wurlod** » Thu Jun 15, 2006 2:16 am

APT_PM_SHOW_PIDS will cause player processes to log a message on startup indicating their process IDs.

APT_PM_NO_SHARED_MEMORY can force player processes to use TCP/IP sockets instead of shared memory for inter-process communication on an SMP environment.

You can restrict memory usage in the Sort stage and in input link sorting.

Research other techniques in the Parallel Job Advanced Developer's Guide