[GE users] interval of time between job submission

reuti reuti at staff.uni-marburg.de
Fri Jul 30 10:13:10 BST 2010


    [ The following text is in the "utf-8" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some characters may be displayed incorrectly. ]

Hi,

Am 29.07.2010 um 17:27 schrieb lemaitreh:

> Sorry, I have a very basic question. I noticed than my SGE run out of memory when I use all slots available, so I reduced the number of slots to make it work. Monitoring my jobs, I noticed that they use a lot of memory at the beginning and get back to normal after a period of time.

varying resource requests over the runtime of the job are not available. So any fixed resource request won't help per se.


> Another solution instead of reducing the number of slots would be to put an interval of time between each job submission to avoid they reach the maximum at the same time. Could you tell me how to set this option properly? I tried to look in the scheduler but I did not find the correct parameters.

But you can implement to request the amount of memory you need after the initial startup of the job by using virtual_free as this is not enforced (otherwise the job would be killed during its first phase where it needs more):

http://gridengine.info/2009/12/01/adding-memory-requirement-awareness-to-the-scheduler

Having done so, the scheduler needs to know, that it needs more during the startup of a job. This can be done by a "load_adjustment" in the scheduler configuration (and putting a really hight value there) `qconf -msconf` and a suitable setting for its decrement in "load_adjustment_decay_time".

-- Reuti


> Thanks,
>  
> Hervé
>  
> `·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·?
> Hervé Lemaître
> U1000 "Imagerie et Psychiatrie"
> INSERM - CEA - Faculté de Médecine Paris Sud 11
> Service Hospitalier Frédéric Joliot
> 4, Place du Général Leclerc
> 91401 ORSAY, FRANCE
> Tél:  (+33) 1 69 86 77 84
> Fax: (+33) 1 69 86 78 10
> `·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·???``·.??.·?
>

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=271068

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



More information about the gridengine-users mailing list