[torqueusers] [torquedev] TORQUE 2.5.4 released

Ken Nielson knielson at adaptivecomputing.com
Tue Dec 28 11:21:29 MST 2010


On 12/28/2010 11:14 AM, Steve Crusan wrote:
> Nice work!
>
>
> Just to be clear, at the moment, one cannot set an attribute in the 
> queue configuration (or the server's nodes file) for GPUs, but only in 
> the node's mom_config file? I couldn't find anything in the source to 
> support something similar to setting resource attributes for a node: 
>       set node bh001 gpus = $
Currently, the number of GPUs can only be set in the 
TORQUE_HOME/server_priv nodes file. Nothing is currently available to do 
a configuration on the MOM node.
>
> As for the second part about qrun (this may be a question for the maui 
> list...), is it necessary to use the qrun to manually run a job with a 
> nodes=$:ppn=$:gpu=$ in the resource listing? I was under the 
> impression that using GRES (in Maui) could support this type of 
> functionality, at least in a limited way.
GRES is separate from the nodes=$:ppn=$:gpu=$. GPUs can still be 
scheduled as a GRES and the :gpu=$ syntax not used. If you use the new 
syntax Moab 5.x will not run the job. It does not know what to do with 
it. I have not tried Maui but I expect it will do the same thing.
>
> Sorry rather basic questions, we've just gotten some GPUs in, and 
> we're working to integrate them with the rest of our cluster as 
> seamlessly as possible.

Who is the vendor of your GPU? We plan to have auto-detection support 
for NVIDIA soon.

>
> Thanks!
>
> ~Steve
>
>
> On 12/28/10 12:45 PM, "Ken Nielson" <knielson at adaptivecomputing.com> 
> wrote:
>
>       It needs to be noted that currently, only the Moab 6.0 scheduler
>     is able to schedule jobs using GPU support.  If jobs are submitted
>     with schedulers other than Moab 6.0 the jobs can be executed using
>     the qrun command.
>
>      Ken Nielson
>      Adaptive Computing
>
>
>      On 12/28/2010 10:24 AM, Ken Nielson wrote:
>
>          TORQUE 2.5.4 has been released.
>
>          GPGPU support is a new feature that has been added to this
>         release. GPUs can be allocated and consumed similar to the way
>         np (processors) are allocated and consumed. The current
>         support is configured in the $TORQUE_HOME/server_priv/nodes
>         file. Future versions of TORQUE will add auto-detection of
>         GPUs plus report statistics for GPUs. For more information
>         about how to use GPU support in Torque visit
>         http://www.clusterresources.com/products/torque/docs/1.5nodeconfig.shtml
>          and also section 2.1.2 at
>         http://www.clusterresources.com/products/torque/docs/2.1jobsubmission.shtml#resources
>
>
>          Several bug fixes were made to this version of TORQUE. See
>         the CHANGELOG in the distribution to review these fixes.
>
>          The tarball can be downloaded from
>         http://www.clusterresources.com/downloads/torque/torque-2.5.4.tar.gz
>         <http://www.clusterresources.com/downloads/torque/torque-2.5.4.tar.gz>
>
>
>          Thanks to everyone who helped get this build ready for release.
>
>          Ken Nielson
>          Adaptive Computing
>
>
>
>         _______________________________________________
>         torquedev mailing list
>         torquedev at supercluster.org
>         http://www.supercluster.org/mailman/listinfo/torquedev
>
>
>
>     ------------------------------------------------------------------------
>     _______________________________________________
>     torqueusers mailing list
>     torqueusers at supercluster.org
>     http://www.supercluster.org/mailman/listinfo/torqueusers
>
>
>
>
> ----------------------
> Steve Crusan
> System Administrator
> Center for Research Computing
> University of Rochester
> https://www.crc.rochester.edu/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torqueusers/attachments/20101228/4f2ead50/attachment-0001.html 


More information about the torqueusers mailing list