Restart QZDASOINIT Jobs

rhunt

Well Known Member
I have a lot of activity (OneWorld, Reporting, Adhoc Analysis) running through ODBC. I often accumulate about 40GB of temporary storage in QZDASOINIT. Does anyone know the best way to stop and restart these jobs every so often? Would it be to stop the QUSRWRK susbsystem and restart it? If so, are there any dependencies that need to be stopped and restarted at the same time?
 
Also, as an extention to this question... Is there a way to set a max memory limit, say 3GB on each QZDASOINIT jobs and then enforce a predefined behavior once the limit has been reached. For instance, once a 3GB threshold has been reached, continue to use whatever memory is required to finish the active request. Once complete, the job should end itself and restart a new QZDASOINIT job to replace it.

Any way to do this?

TIA

Ryan Hunt
 
Ryan,

I'm not in front of my 400 at the moment. But on the PJ job entry in the SBS you can change the number of time the job can be re-used. If you change this to a low number the jobs will end more often, thus cleaning up the temporary space. My own experience is that the amount of temp space is really related to the type of SQL's that are generated by OW. Sometimes JDE creates some pretty ugly joins that end up being implemented with temp files, these take up a lot of space.

I don't know of a way to limit the amount of TEMP storage a job uses. I also limit the amount of CPU time the jobs can use to 5000000 MS, if the job uses more than this I assume it is a run away.

Hope this helps.

Tom Davidson
 
Tom, thanks for the response. I currently have my QZDASOINIT jobs set at a prestart of 600. At threshold of 10 left available, 50 more start. Once a thread has been used 200 times, it is ended. Do you think the 200 is too many?
 
Ryan,

I have mine set to initial #=1700; Threshold 10; Additional 100; max uses: 100.

I also have my qpwfserver class set as follows: runpty 20; timeslice 3000; max cpu 5000000;

finally I run my QZDASOINIT jobs is a separate pool with 7G RAM and max act of 110. I run about 2000 active jobs 19 hours a day. My low is normally about 900.

Hope this helps.

Tom Davidson
 
If you restart the QUSRWRK subsystem you also need to strhostsvr for client access, I do believe.

The temp space is regulated by the class associated with teh job. Look at the prestart job definition for the class. I created a new class and then assigned the values I wanted.
 
Jean,

Thanks for the tip. I never noticed the option on the class before.

Tom Davidson
 
Back
Top