[torqueusers] Restrict queue to group?

Peter Smith peter.smith3882100 at gmail.com
Fri Apr 9 04:16:53 MDT 2010


I have tried to configure acl_groups and read pbs_queue_attributes,
when a user that is not member of my group now tries to submit to the
restricted queue, a nice error/warning message that says "ACL is not
satisfied ..." appears, this is great.

Thanks.

On Thu, Apr 8, 2010 at 4:47 PM, Coyle, James J [ITACD] <jjc at iastate.edu> wrote:
> Peter,
>
>  The following describes how you can accomplish what you want with pbs_sched.
>
> Restricting a queue to a group or groups:
> -------------------------------------------
>   Yes, you can restrict a queue to a group or set of groups
> using the acl_groups attribute on a queue.
>
>   E.g. I have a queue named "lq" which is restricted to only two
> Groups: ccresearch and an_sci2
> via the qmgr commands:
>
> set queue lq acl_groups = ccresearch
> set queue lq acl_groups += an_sci2
>
>   The groups need to explicitly request this queue via qsub -q lq  ...
> I was not able to set this up within my default routing queue.  If you
> find a way, let me know.
>
>
> Getting jobs from this group to run ahead of other groups:
> -----------------------------------------------------------
>   To get jobs run immediately when a job comes in, you need nodes idle
> except when a job from that queue is running.
>  I have this requirement for small short jobs on the cluster for students,
> as that is the primary pur[pose of the cluster, and most jobs are class
> assignments which can at least be debugged in one hour run on 4 nodes.
>  To implement this I set up a queue named "small_short" which allows only 1 hour
> runs on at most 4 nodes.  I arrange the nodect.available for the other queues
> so that when they are added up, there are at least 4 nodes available.
> I have nodect.available set up for small_short so that many of these jobs can run
> at once with no more that one per user at a time.
>
>  This might seem wasteful of nodes, but the primary purpose of the cluster is
> fast access for short classroom assignments.  Keeping the cluster busy all the
> time is secondary. (After all, what good is a parallel program that solves
> your problem in 15 minutes after a wait in the queue of 48 hours, if you
> could have solved the same problem on a desktop in 4 hours.)
>
> Good luck on our setup,
> - Jim C.
>
>
>  James Coyle, PhD
>  High Performance Computing Group
>  115 Durham Center
>  Iowa State Univ.           phone: (515)-294-2099
>  Ames, Iowa 50011           web: http://www.public.iastate.edu/~jjc
>
>
>
> -----Original Message-----
> From: torqueusers-bounces at supercluster.org [mailto:torqueusers-bounces at supercluster.org] On Behalf Of Kevin Van Workum
> Sent: Thursday, April 08, 2010 7:42 AM
> To: Torque Users Mailing List
> Subject: Re: [torqueusers] Restrict queue to group?
>
> On Thu, Apr 8, 2010 at 2:38 AM, Josh Bernstein
> <jbernstein at penguincomputing.com> wrote:
>> Hi Peter,
>>
>> I don't think that's possible without using Maui. There is always the
>> method of creating a wrapper script around qsub to route your jobs to
>> the proper queue.
>>
>> -Josh
>>
>> On Apr 7, 2010, at 11:27 PM, "Peter Smith"
>> <peter.smith3882100 at gmail.com> wrote:
>>
>>> Hello
>>>
>>> I have been experimenting with creating multiple queues in Torque. For
>>> different reasons i am restricted to use the default scheduler
>>> pbs_sched. Is is possible to restrict one queue to a specific group
>>> [the group is a normal Unix group created on all nodes] and execute
>>> the jobs in that queue before jobs in the standard queue, even if the
>>> jobs in the standard queue was submitted first?
>
> For restricting queues by groups, take a look at the
> pbs_queue_attributes manpage. I believe that acl_groups is what your
> looking for.
>
>>> _______________________________________________
>>> torqueusers mailing list
>>> torqueusers at supercluster.org
>>> http://www.supercluster.org/mailman/listinfo/torqueusers
>> _______________________________________________
>> torqueusers mailing list
>> torqueusers at supercluster.org
>> http://www.supercluster.org/mailman/listinfo/torqueusers
>>
>
>
>
> --
> Kevin Van Workum, PhD
> Sabalcore Computing Inc.
> Run your code on 500 processors.
> Sign up for a free trial account.
> www.sabalcore.com
> 877-492-8027 ext. 11
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers
>


More information about the torqueusers mailing list