[Mauiusers] problem with number job queued jobs

Bas van der Vlies basv at sara.nl
Mon Jun 7 03:14:35 MDT 2010


On 04-06-10 15:54, Giacinto Donvito wrote:
> Hi All,
>
> our maui installation is experiencing this problem:
>
> ###########################################################
> Total Jobs: 3553 Active Jobs: 793 Idle Jobs: 0 Blocked Jobs: 2760
>
> NOTE: output truncated
> ###########################################################
>
> I would like to ask if it is possible to change the limit in maui that
> block the daemon to look at the job over the "3553" ?
>
> Could I recompile maui with different parameters in order to avoid that?
>
> Or could I change the code somewhere to avoid that?
>
> This is causing a lot of problems in scheduling the jobs, as usually our
> cluster has few thousands of jobs in queue.
>
> Thank you to any comment or help that you could provide me.
>
> Cheers,
> Giacinto
>

What we have done is limiting the number of jobs that an user can submit in 
an execution queue, eg: (200 node cluster):
  * we have two execution queues just for administration:
    * q_serial   -->  set queue q_parallel max_user_queuable = 120
    * q_parallel -->  set queue q_parallel max_user_queuable = 60

  * batch queue, is a routing queue.

When one user submits more then 120 jobs for the serial queue. Torque will 
hold the jobs in the batch queue which is a routing queue. Maui does not 
process jobs that are in a routing queue. With this setup users can submit 
a lot of jobs and maui still process the jobs.

Regards


-- 
********************************************************************
*  Bas van der Vlies                    e-mail: basv at sara.nl       *
*  SARA - Academic Computing Services   Amsterdam, The Netherlands *
********************************************************************


More information about the mauiusers mailing list