[gridengine users] fair share for many single and one parallel

[Reuti]

Am 03.03.2011 um 23:10 schrieb Gavin W. Burris:

> Reuti,
> 
> I have enabled reservations per the blog post, but it does not seem to
> help under full utilization.  "qstat -g c" never shows reserved slots.

Correct. This column shows advance reservation, not the resource reservation. Resource reservation can be anything besides slots.


> Notice that jobs submitted after the parallel jobs are still filling in
> freed slots:
> 
> 65034 0.50000 search120. user1 r  03/03/2011 14:59:01 all.q at node23  1
> 65035 0.50000 search121. user1 r  03/03/2011 15:09:46 all.q at node28  1
> 65036 0.50000 search122. user1 r  03/03/2011 15:15:01 all.q at node29  1
> 65072 0.50079 cl_15000   user2 r  03/03/2011 15:30:16 all.q at node14  1
> 65073 0.50079 cl_10000   user2 r  03/03/2011 16:14:01 all.q at node15  1
> 65074 0.50079 cl_09000   user2 r  03/03/2011 16:38:16 all.q at node23  1
> 65075 0.50079 cl_08000   user2 r  03/03/2011 16:57:31 all.q at node09  1
> 65080 0.51000 r2617.run. user3 r  03/03/2011 15:32:46 all.q at node09  1
> 65094 0.51000 r1853.run. user3 r  03/03/2011 16:47:16 all.q at node23  1
> 64631 0.60250 CM2Mc_Cont user4 qw 03/01/2011 17:20:12              44
> 63206 0.55949 om1p7_spin user4 qw 02/25/2011 13:43:34              24
> 65071 0.53988 test       user5 qw 03/03/2011 14:43:34              16
> 65093 0.52343 test       user5 qw 03/03/2011 16:40:07              10
> 65037 0.00000 search123. user1 qw 03/03/2011 11:59:20               1
> 65038 0.00000 search124. user1 qw 03/03/2011 11:59:20               1
> 65039 0.00000 search125. user1 qw 03/03/2011 11:59:20               1

Mmh - do you request h_rt for all job?

-- Reuti


> Cheers.
> 
> On 02/16/2011 09:48 AM, Reuti wrote:
>> Am 16.02.2011 um 15:37 schrieb Gavin W. Burris:
>> 
>>> I have a queue configured for user fair share.  There are always many
>>> single-slot jobs waiting in the queue.  I have noticed that a 24-slot
>>> parallel job has been stuck in qw for many days now, with
>>> later-submitted single-slot jobs filling in all available slots ahead of it.
>>> 
>>> With single-slot jobs always waiting in the queue, there will never be
>>> 24 free slots for the parallel job.  Is there a recommended solution to
>>> this predicament?
>> 
>> This will help:
>> 
>> http://gridengine.info/2006/05/31/
>> 
>> -- Reuti
>> 
>> 
>>> Thanks,
>>> -- 
>>> Gavin W. Burris
>>> Senior Systems Programmer
>>> Information Security and Unix Systems
>>> School of Arts and Sciences
>>> University of Pennsylvania
>>> _______________________________________________
>>> users mailing list
>>> users at gridengine.org
>>> https://gridengine.org/mailman/listinfo/users
>> 
>> 
> 
> -- 
> Gavin W. Burris
> Senior Systems Programmer
> Information Security and Unix Systems
> School of Arts and Sciences
> University of Pennsylvania