Page 1 of 1

8 node configuration lead fails in parallel execution of DS

Posted: Thu Mar 24, 2011 3:10 am
by Sgiri1
Hi All,

I referred to several posts posted earlier on the below error message:

APT_PMPlayer::APT_PMPlayer: fork() failed, Resource temporarily unavailable

I have tried all suggested option like

1)set Maximum number of PROCESSES allowed per user to a higher value: This parameter already has 2048 set to it which is way too high.

2) has altered the value for $APT_MONITOR_SIZE and APT_MONITOR_TIME to 100000 and 25.

3)has set $APT_NO_JOBMON = TRUE

None of the above options seem to work. When i alter the configuration file to 4 nodes it seem to work fine. But our cpu configuration can support 8 nodes and i am unable to use it. I get the error when i try to run more than 4 parallel jobs at the same time with 8 node configuration. When treid to run one after the other i don't face this issue.

Any suggestion will be of great help.

Posted: Thu Mar 24, 2011 8:12 am
by jwiles
Regarding comment #1: Why do you believe that 2048 is way too high? With IS, you're not running a single-process application. The usage model is quite different than that for a desktop or database server.

Are all of your jobs run using the same userid, such as dsadm, or is each job potentially run using a different userid? If everything is run using one id, the process count for that user can easily hit the maximum you've set when you're running multiple jobs in multiple degrees of parallelism. Use your system tools to monitor the total number of processes active per user while jobs are starting. I've seen that 2048 value get blown out of the water several times.

How complex is the job (or jobs) you're trying to run? Set $APT_DUMP_SCORE=1 to see just how many processes it attempts to start at the degree of parallelism you're running.

Regards,