[GE users] nice values and sge 6.0

Andreas Haas Andreas.Haas at Sun.COM
Tue Jul 20 09:52:59 BST 2004


In 6.0 the 5.3 NO_REPRIORITIZATION execd_param is available
as sched_conf(5) reprioritize_interval. Currently
reprioritize_interval default is non-zero which causes
queue_conf(5) priority be set though at job startup but soon
after be changed.

To effectuate queue priority values remain after job startup
unchanged just set reprioritize_interval to 0. We're changing
the installation default to 0 to overcome that pitfall. See
also #1139.

Andreas



On Mon, 19 Jul 2004, Don Shesnicky wrote:

>
> I had spotted this email from Andy in the archives but it seemed to
> imply that
> REPRIORITIZATION was different than nice:
>
>  Nice and priority are two different things. The priority is derived
> from the
>  CPU time used (probably in the last *kernel* scheduling interval).
>  With SGE/SGEEE 5.3beta2 and SGE/SGEEE 5.3 you can configure the full
> nice
>  range form -20 (high) to 20 (low) in the queue configuration (through
> the
>  "priority" parameter.
>  If you are using SGEEE by default the scheduler dynamically
> reprioritizes
>  the running jobs (in an OS dependent range).
>  If you don't want SGEEE to reprioritize running jobs set in the global
>  cluster config    qmaster_params NO_REPRIORITIZATION=1
>
> I'll research the qconf command for this setting as that is what
> sge_conf
> points to but off hand do you know how to set it on a particular group
> of
> hosts within a queue. As I mentioned on a different thread I have two
> groups
> of hosts in my d.reg queue - @digital and @analog and only want this
> setting
> on the analog hosts.
>
> I beginning to think that the overhead of moving to the command line
> interface
> is worth it as the qui doesn't seem to cut it.
>
> Don
>
> -----Original Message-----
> From: Chris Dagdigian [mailto:dag at sonsorol.org]
> Sent: Monday, July 19, 2004 5:10 PM
> To: users at gridengine.sunsource.net
> Subject: Re: [GE users] nice values and sge 6.0
>
>
> I could be totally wrong, but I think this is what is going on:
>
> If you really want nice values you need to turn off "dynamic
> reprioritization" within N1GE/SGE 6.
>
>  From what I recall there are 2 things you need to change
>
> (1) execd_param NO_REPRIORITIZATION=true (sge_conf manpage) at queue
> level
>
> (2) set REPRIORITIZE_INTERVAL to 00:00:00 (sched_conf manpage) within
> the scheduler configuration
>
>
> After doing that you should see your queue level nice values start to be
>
> expressed I belive.
>
> Regards,
> Chris
>
>
>
>
> Don Shesnicky wrote:
>
> >
> > I have three queues setup in sge 6.0, a high, medium and low where I
> set
> > different nice
> > values via the qmon queue window. I've tried a number of different
> > settings both plus and
> > minus and never see any unix nice value on the processes via both top
> or
> > ps. Should I?
> >
> > Don
> >
>
> --
> Chris Dagdigian, <dag at sonsorol.org>
> BioTeam  - Independent life science IT & informatics consulting
> Office: 617-665-6088, Mobile: 617-877-5498, Fax: 425-699-0193
> PGP KeyID: 83D4310E iChat/AIM: bioteamdag  Web: http://bioteam.net
>
> ---------------------------------------------------------------------
> 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
>
>

---------------------------------------------------------------------
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