[torqueusers] How to stop the submission for new jobs (for maintenance)
scrusan at UR.Rochester.edu
Fri Oct 1 08:09:21 MDT 2010
Yes, the reservation in Maui is the best way to go. By just shutting down the scheduler, no small/interactive/etc type jobs will start(as stated before), so it's kind of like you're extending your downtime if you just shut your scheduler down until the actual maintenance window.
Also, I don't think the OP is using the torque scheduler, so I don't see how stopping it would work. I think the qmgr -c 'set server scheduling = False' only matters if one is using the torque scheduler.
Center for Research Computing
From: torqueusers-bounces at supercluster.org on behalf of Lawrence Lowe
Sent: Fri 10/1/2010 8:47 AM
To: Torque Users Mailing List
Subject: Re: [torqueusers] How to stop the submission for new jobs (for maintenance)
Someone will probably correct me if I'm wrong, but I think the "set server
scheduling" just determines whether the pbs_server kicks the scheduler
immediately a job comes in. But the scheduler has its own polling cycle
and will pick up new jobs anyway.
So I guess it's "service maui stop", or put in a reservation. A
reservation is better because then short jobs will continue to get
scheduled provided their walltime indicates they will finish before the
maintenance period. Anyway, that's what we do!
Tel: 0121 414 4621 Fax: 0121 414 6709 Email: L.S.Lowe at bham.ac.uk
On Fri, 1 Oct 2010, giggzounet wrote:
> Le 01/10/2010 14:28, Grid-Admins a écrit :
>>> you mean disable maui ? or enter qmgr -c "set server scheduling = False" ?
>> I guess it's the same. On our head node I ran:
>> /etc/init.d/torque-scheduler stop
>> The job submissions are managed by the pbs server. As long as no
>> scheduler sends any requests the jobs will stay in the queue(s).
>> So either you tell your server to ignore scheduling requests or you stop
>> sending scheduling requests at all.
> maui running and qmgr -c "set server scheduling = False" doesn't work. a
> new job starts.
> torqueusers mailing list
> torqueusers at supercluster.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the torqueusers