One correction, it is "resources_default" (with a "s"), NOT
resource_default.
Hi Wei,
Just another suggestion, that we are using here:
qmgr -c "s q atlas resource_default.neednodes=atlas"
qmgr -c "s q biomed resource_default.neednodes=biomed"
...
qmgr -c "s q dteam resource_default.neednodes=dteam"
AND then, for each node that would map to a certain queue:
qmgr -c "s n <node name> properties+=atlas"
qmgr -c "s n <same node name> properties+=biomed"
...
As a result you may map nodes to queues and give a fair share to each
public/local VO.
(a node may have a list of properties; a queue would use that node if
its "neednodes" matches one of node's properties.)
Regards,
Dan
Wei Xing wrote:
> Hi, Steve & Jan,
>
> Thanks a lot for your suggestions.
>
> I just fixed the bug in the /opt/lcg/libexec/lcg-info-dynamic-pbs. And
> I will contact with biomed people for this.
>
> Thanks again,
>
> Cheers
>
> Wei
>
>
>
>
>
> Steve Traylen wrote:
>
>> On Thu, Jul 14, 2005 at 02:11:37PM +0200 or thereabouts, Jan Astalos
>> wrote:
>>
>>
>>> Wei Xing wrote:
>>>
>>>
>>>> Dear all,
>>>>
>>>
>>> Hi Wei,
>>>
>>>
>>>
>>>> We are now participate BioMed DC.
>>>>
>>>> We have together 13 WNs, which 6 dedicate BioMed DC to running
>>>> BioMed jobs. However, since we have other 7 WNs free, RB sends us
>>>> BioMed jobs continuously (so far we have 164 jobs). How can we
>>>> let RB know we are now overload?
>>>>
>>>
>>
>> Hi Wei,
>>
>> I think you should ask the biomed people why they are doing it.
>> [log in to unmask]
>> or you can ask them via the
>> https://cic.in2p3.fr/index.php?id=rc&subid=rc_publish&js_status=2
>>
>> which I used yesterday to contact them and worked.
>>
>> One problem they are facing is that many of the larger sites
>> including RAL
>> are unable to run their work. (or rather they are unable to run their
>> jobs at the site)
>>
>> Steve
>>
>>
>>> You can set a limit to GlueCEPolicyMaxTotalJobs for biomed. Set in qmgr
>>>
>>
>>
>>
>>
>>
>>> set queue biomed max_queuable = 10
>>>
>>> and fix the bug in /opt/lcg/libexec/lcg-info-dynamic-pbs by
>>> changing max_queueable to max_queuable. After a while your BDII
>>> should report for biomed:
>>>
>>> GlueCEPolicyMaxTotalJobs: 10
>>>
>>> But before that you should put the biomed queue to 'Draining' state
>>> to get rid of queued jobs by:
>>>
>>> set queue biomed enabled = False
>>>
>>> Cheers,
>>>
>>> Jan
>>>
>>>
>>>
>>>
>>>> Thx,
>>>>
>>>> Wei
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>> --
>>> Jan Astalos
>>> Institute of Informatics, Slovak Academy of Sciences
>>>
>>
>>
>>
>>
>
>
|