Re: Automatically killing long running jobs

Jim.McLean wrote:

I'm looking for a solution to my problem of long running reports. Everyone
once in a while a user gets carried away and submits a very, very large
report to the system. What I would like to do is kill any reports that run
for more than X minutes. These reports run in our reporting
queue/subsystem, so it would be safe to kill any job in that subsystem
that's been running for more that X minutes.

Any ideas on how best to accomplish this? Or am I off the mark in trying
this?

Jim:

The "best" answer _might_ depend on what the problem is with long-running jobs.

Is the problem that other jobs wait too long? Maybe all that's needed is multiple jobqs or multi-threaded jobqs. Or maybe a more automated way to service waiting jobs would be useful.

Is the problem that reports get to be too long (too many printed lines)? Maybe the MAXRCDS() attribute needs adjustment on the printer file.

Is the problem that queries create temporary tables that are too big or that temporary indexing takes too much of system resources? Maybe various limits such as MAXTMPSTG() on your job classes or MAXSTG() on user profiles need adjustment.

And maybe other things need to be done for any of those or for other conditions.

Can you describe the conditions that you want to resolve more fully? The "best" way to fix what's wrong begins with being sure we know what's wrong.

I never like killing long-running jobs as long as the jobs are working as expected. Once killed, all of the resources that had been used are sunk costs that are gone. Sometimes it works just to have those jobs changed to minimize their impact on everything else but allow them to plod on to their end. (Sometimes.)

Tom Liotta