[GE users] Ressource quota

Reuti reuti at staff.uni-marburg.de
Wed Aug 22 11:04:10 BST 2007


    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

Hi,

Am 22.08.2007 um 11:36 schrieb Jean-Paul Minet:

> We have a cluster composed of 128 biproc nodes and a few SMP  
> machines (8 cpus) member of the @smp group.  Using the new  
> ressource quota feature of SGE 6.1, we wanted to limit the total of  
> cpu used by any user to 64, among which 16 at most could be used on  
> the SMP machines.  Here is the rule we defined:
>
> {
>   name         max_slots
>   description  Maximum slots per user
>   enabled      TRUE
>   limit        users {*} hosts @smp to slots=16
>   limit        users {*} to slots=64
> }
>
> The limits seem to be enforced:
>
>
>> qquota -u sdubois
> resource quota rule limit                filter
> ---------------------------------------------------------------------- 
> ----------
> max_slots/1        slots=16/16          users sdubois hosts @smp
> max_slots/2        slots=56/64          users sdubois

limits will be checked in order in which they appear in the resource  
quota definition until a valid one is found, then the job will run.  
If you want to have both limits in effect, you will have to split  
this to two resource quota definitions.

-- Reuti


> but the maximum 64 total doesn't include the SMP machines (lmexec-6  
> and lmexec-7 are members of @smp):
>
>> qstat -u sdubois
>  78102 0.00019 graph      sdubois      r     08/19/2007 05:34:54  
> all.q at lmexec-110                   8         78099 0.00019  
> graph      sdubois      r     08/18/2007 10:20:58  
> all.q at lmexec-37                    8         78210 0.00019  
> graph      sdubois      r     08/19/2007 18:51:27  
> all.q at lmexec-50                    8         78563 0.00019  
> graph      sdubois      r     08/21/2007 14:53:31  
> all.q at lmexec-57                    8         78107 0.00019  
> graph      sdubois      r     08/19/2007 13:55:49  
> all.q at lmexec-6                     8         78543 0.00019  
> alk8       sdubois      r     08/21/2007 08:54:50  
> all.q at lmexec-7                     4         78766 0.00019  
> alk8       sdubois      r     08/22/2007 09:16:14  
> all.q at lmexec-7                     4         78501 0.00019  
> alc8       sdubois      r     08/21/2007 03:51:35  
> all.q at lmexec-74                    8         78341 0.00019  
> alc8       sdubois      r     08/20/2007 10:12:15  
> all.q at lmexec-85                    8         78362 0.00019  
> alk8       sdubois      r     08/20/2007 12:11:11  
> all.q at lmexec-92                    8
> I am a bit puzzled as the line
>
>   limit        users {*} to slots=64
>
> isn't referring to any particular host group and therefore should  
> be calculated accross the whole cluster, thus including the @smp  
> group.
>
> Any insight ?
>
> Thanks for any help
>
> Rgds
>
> -- 
> Jean-Paul Minet
> Institut de Calcul Intensif et de Stockage de Masse (CISM)
> Université Catholique de Louvain
> http://www.cism.ucl.ac.be
>
> Place Croix du Sud, 1
> 1348 Louvain-la-Neuve
> Belgium
>
> Tel: (32) 010/47.35.67
> Fax: (32) 010/47.34.52
>
> ---------------------------------------------------------------------
> 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