[GE users] Resource Quota with parallel environment on 6.2u5

sumyee sum-yee.lai at hp.com
Fri Apr 30 17:49:46 BST 2010


Thanks everyone!

Using -q high.q at tester works!

Sum Yee

-----Original Message-----
From: icaci [mailto:hristo at mc.phys.uni-sofia.bg] 
Sent: Tuesday, April 20, 2010 12:27 AM
To: users at gridengine.sunsource.net
Subject: Re: [GE users] Resource Quota with parallel environment on 6.2u5

Hi,

On 16.04.2010, at 18:31, sumyee wrote:

> Hello,
> 
> I am experiencing some odd behavior with resource quota.  I have defined the following quota to limit the number of slots a user can consume in high.q:
> 
> {
>  name    max_high_slots
>  description "limit number of slots in high priority queue per user"
>  enabled  TRUE
>  limit  users {*} queues high.q to slots=24
> }
> 
> It normally runs fine, unless I have a hard request to go to a specific host AND a request to run in parallel environment.
> 
> Job runs in the following situations:
>> qsub -q high.q sleep.sh
>> qsub -q high.q -l h=tester sleep.sh
>> qsub -q high.q -soft -l h=tester sleep.sh
>> qsub -q high.q -pe smp 2-4 sleep.sh
> 
> The job doesn't run when requested like this:
>> qsub -q high.q -l h=tester -pe smp 2-4 sleep.sh
> 

This looks and feels like the problem we encountered when using RQS to limit slot usage in some parallel queues:
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=247462

It is definitely a bug in the SGE scheduler code that sorts and screens suitable queues/hosts for parallel tasks and is unrelated to the queue sort method that you've set in the scheduler config.

Hristo

> 
>> qquota
>> resource quota rule limit        filter
>> ---------------------------------------------------
>> max_high_slots/1    slots=9/24   users sumyee queues high.q
> 
> Scheduler reports the following:  cannot run because it exceeds limit "sumyee////" in rule "max_high_slots/1"
> 
> It doesn't seem to matter whether I have the scheduler sort by seq_no or by load.  I get the same result.  Is this issue related to the way the RQS is defined?  Is there a way around it?
> 
> Thanks,
> Sum Yee
> 
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=253685
> 
> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].

--
Dr Hristo Iliev
Monte Carlo research group
Faculty of Physics, University of Sofia
5 James Bourchier blvd, 1164 Sofia, Bulgaria
http://cluster.phys.uni-sofia.bg/hristo/

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=254131

To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=255570

To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].



More information about the gridengine-users mailing list