[GE users] h_fsize and mpi

reuti reuti at staff.uni-marburg.de
Wed Dec 16 12:42:30 GMT 2009


Hi,

Am 16.12.2009 um 13:27 schrieb pablorey:

>     Hi Reuti,
>
>     I have done the test that you suggested and it seems that it  
> confirm my thesis.
>
>     I have reserved a node with h_fsize=400G and I done the  
> following tests (mpi_4p is configure to use 4 slots per node):
>     1) qsub h_fsize=100G -pe mpi_4p 4: The job ran in the reserved  
> nodo and the file size limit was set to 100G.
>     2) qsub h_fsize=200G -pe mpi_4p 4: The job ran in other node  
> with h_fsize=900G and the file size limit was set to 100G.

typo: "... was set to 200G" you meant?

>     3) qsub h_fsize=400G -pe mpi_4p 4: The job was not enqueued  
> because we don't have nodes with h_fsize>=1600G.

Did you make h_fsize consumable? Otherwise SGE shouldn't reject it I  
think.


>     4) qsub h_fsize=400G -pe mpi 4: The job ran in 3 nodes (2 of  
> them with h_fsize=400G and other with h_fsize=900G) and the file  
> size limit was set to 100G.

typo: "... was set to 400G" you meant?

-- Reuti


>
>     So we can deduced that as you said the limit is per process,  
> and that GE do the same as it is verified in the case 3) and 4).
>
>     Thanks by your suggestion,
>     Pablo
>
>
>
> On 16/12/2009 11:43, reuti wrote:
>>
>> Hi, Am 16.12.2009 um 11:09 schrieb pablorey:
>>>
>>> Hello, I have a doubt about h_fsize and MPI jobs. In a MPI job,  
>>> the h_fsize is limited by slot or limited by node?. I know that  
>>> the s_vmen value is the limit for each slot. For example, if a  
>>> sent the following job: qsub -l  
>>> num_proc=1,s_rt=10:00:00,s_vmem=2G,h_fsize=20G -pe mpi 32 job.sh
>> like all system limits it's per process. To check it, you submit a  
>> job with a script like: #!/bin/sh ulimit -aH echo ulimit -aS to  
>> get the set limits output. Only limits handled by SGE are  
>> different (man queue_conf), i.e. h/ s_cpu and h/s_vmem. -- Reuti
>>>
>>> the memory limit by each slot is established to 2G, so in each  
>>> node the memory limit is set to "2G * number_of_assigned_slots".  
>>> What about the h_fsize limit?. Is similar to s_vmen or it is set  
>>> by node?. That is, in the previous example, the limit is 20G or  
>>> "20G * number_of_assigned_slots". Regards. -- Pablo Rey Mayo  
>>> Tecnico de Sistemas Centro de Supercomputacion de Galicia (CESGA)  
>>> Avda. de Vigo s/n (Campus Sur) 15705 Santiago de Compostela  
>>> (Spain) Tel: +34 981 56 98 10 ext. 233; Fax: +34 981 59 46 16  
>>> email: prey at cesga.es; http://www.cesga.es/  
>>> ------------------------------------------------ NOTA: Este  
>>> mensaje ha sido redactado intencionadamente sin utilizar acentos  
>>> ni caracteres especiales, para que pueda ser visualizado  
>>> correctamente desde cualquier cliente de correo y sistema.  
>>> ------------------------------------------------ <xacobeo.jpg>
>> ------------------------------------------------------ http:// 
>> gridengine.sunsource.net/ds/viewMessage.do? 
>> dsForumId=38&dsMessageId=233677 To unsubscribe from this  
>> discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].
>
> -- 
> Pablo Rey Mayo
> Tecnico de Sistemas
> Centro de Supercomputacion de Galicia (CESGA)
> Avda. de Vigo s/n (Campus Sur)
> 15705 Santiago de Compostela (Spain)
> Tel: +34 981 56 98 10 ext. 233; Fax: +34 981 59 46 16
> email: prey at cesga.es; http://www.cesga.es/
> ------------------------------------------------
> NOTA: Este mensaje ha sido redactado intencionadamente sin utilizar
> acentos ni caracteres especiales, para que pueda ser visualizado
> correctamente desde cualquier cliente de correo y sistema.
> ------------------------------------------------
> <xacobeo.jpg>

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=233698

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



More information about the gridengine-users mailing list