[GE users] Queue Management

templedf dan.templeton at sun.com
Tue Nov 10 15:00:50 GMT 2009


The quick fix is to set job_load_adjustments to NONE in the scheduler 
config.  What Mark's suggesting may not solve the problem.  In short, 
the scheduler picks the least loaded host and then the lowest numbered 
queue for each job.  What Mark's suggesting just reverses that.  The 
real culprit is that every time the scheduler puts a job on a host, it 
adds virtual load to that host, making it no longer the least loaded 
host.  That means the next job it schedules will be directed somewhere 
else.  The amount of load is controlled by the job_load_adjustments.  If 
it's set to NONE, then no load is added, meaning that the least loaded 
host when the first job was scheduled is still the least loaded host for 
the second job, etc.

Daniel


olesen wrote:
>> I have 32 machines with 2 nodes (processors each)
>>
>> I want the queue to do this
>>
>> c1:node1, c1:node2 , c2:node1, c2:node2 ...
>>
>> and so on to allow more space for openMP jobs.
>>
>> At the moment it just seems to put the jobs where ever.
>>     
>
> It is likely sorting by load.
> Take a look at 'qconf -ssconf', you should try
>
>    queue_sort_method   seqno
>
>
> Then it will sort by the machine seqno first and load afterward.
> Your parallel environments might also need adjusting to have 
>
>     allocation_rule    $fill_up
>
>
> /mark
>
>
> This e-mail message and any attachments may contain legally privileged, confidential or proprietary Information, or information otherwise protected by law of EMCON Technologies, its affiliates, or third parties. This notice serves as marking of its "Confidential" status as defined in any confidentiality agreements concerning the sender and recipient. If you are not the intended recipient(s), or the employee or agent responsible for delivery of this message to the intended recipient(s), you are hereby notified that any dissemination, distribution or copying of this e-mail message is strictly prohibited. 
> If you have received this message in error, please immediately notify the sender and delete this e-mail message from your computer.
>
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=226016
>
> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].
>

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=226022

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



More information about the gridengine-users mailing list