[GE users] Job exit status

Beadles, Jeff jeff_beadles at mentorg.com
Fri Apr 2 16:51:14 BST 2004


qacct -j is very, very slow, after the first couple of days of the month.  The reason is that it searches thru the accounting file one line at a time to find the job in question.  We roll our accounting file monthly, and at the end of the month this can take 10+ seconds per job, just to see the exit status.  (We're running ~40k jobs per day on one of the grids here)

You also run into a problem if you use qacct, that if the job finishes just before the log file rolls, and you check for the job status after the roll, it won't report.

What I've done is to have a epilog script that runs after every job.  It gets the exit status of the job from $SGE_JOB_SPOOL_DIR/usage, and stores it in a file outside of grid.
The users run a "grid_job_status $JOBID" to get the exit value from their jobs.

As a result, it takes a few milliseconds to get the results instead of 10 seconds.  When you're waiting for 5,000 jobs to complete, it's a big deal. :-)

(If anyone wants the scripts, email me off-list)

  -Jeff


-----Original Message-----
From: Ron Chen [mailto:ron_chen_123 at yahoo.com] 
Sent: Thursday, April 01, 2004 5:51 PM
To: users at gridengine.sunsource.net
Subject: Re: [GE users] Job exit status

Any reason you couldn't use "qacct"?

 -Ron

--- Keith Rarick <KRarick at hifn.com> wrote:
> Quick question. How do you find the exit status of a job?
> 
> I would expect it to be somewhere in qstat, but I can't find mention 
> of this anywhere in the documentation. The closest thing I've seen is 
> "-t", which prints the exit status of a *task*. But I'm just 
> submitting regular plain old jobs, and when I type "qstat -t -s z" the 
> task exit status column is empty.
> 
> Thanks!
> 
> -Keith
> --
> You cannot make an interface better without making it different.
> If it's a lot better, it will be a lot different.
> 
>     -- Aza Raskin
> 
>
---------------------------------------------------------------------
> To unsubscribe, e-mail:
> users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail:
> users-help at gridengine.sunsource.net
> 


__________________________________
Do you Yahoo!?
Yahoo! Small Business $15K Web Design Giveaway http://promotions.yahoo.com/design_giveaway/

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list