[slurm-users] Complex resource requests for a single job

Christopher W. Harrop christopher.w.harrop at noaa.gov
Wed Jul 25 10:58:38 MDT 2018


Hi Jeff,

That’s what I first looked at, but that page is so confusing I couldn’t decipher if that was indeed what I was looking for.  A SchedMD person replied to me privately and pointed me to the same location.  After a bit of back and forth, it is indeed the heterogeneous job that I am looking for.  But, unfortunately, it turns out that running an executable across multiple components is not supported until version 18.08, which isn’t out yet.  But, I did get my question answered.

So, I’ll have to wait for the next version.  I think it would be helpful to me and others if that page was reworked to describe more clearly what a heterogeneous job is, what a component is, and provide more examples that include basic usage as well as advanced usage.  

Chris
-----------------------------------------------------------------------------------------------------------
Christopher W. Harrop                               email: christopher.w.harrop at noaa.gov <mailto:christopher.w.harrop at noaa.gov>
Global Systems Division                                                      voice: (303) 497-6808 <tel:%28303%29%20497-6808>
NOAA Earth System Research Laboratory                             fax: (303) 497-7259 <tel:%28303%29%20497-7259>
325 Broadway R/GSD6
Boulder, CO 80303






> On Jul 25, 2018, at 10:29 AM, Jeff White <jeff.white at wsu.edu> wrote:
> 
> Sound like you want a heterogeneous job:
> 
> https://slurm.schedmd.com/heterogeneous_jobs.html <https://slurm.schedmd.com/heterogeneous_jobs.html>
> 
> Jeff White
> 
> On 07/24/2018 09:59 AM, Christopher W. Harrop wrote:
>> Hi,
>> 
>> I am sorry if this basic question has been asked before.  I’ve search the documentation and lists but can’t seem to find the answer.
>> 
>> How does one submit a job that requires a non-uniform number of cores per node?
>> 
>> For example, how do you submit a job that needs 1 core on the first node, and 10 cores on each of the next 10 nodes?
>> 
>> In PBS/Torque, this would be accomplished with: "-l nodes=1:ppn=1+10:ppn=10”
>> 
>> This request would be for a single executable that will be run across all the requested resources.
>> 
>> Thanks for your time,
>> 
>> Chris
>> -----------------------------------------------------------------------------------------------------------
>> Christopher W. Harrop                               email: christopher.w.harrop at noaa.gov <mailto:christopher.w.harrop at noaa.gov>
>> Global Systems Division                                                      voice: (303) 497-6808 <tel:%28303%29%20497-6808>
>> NOAA Earth System Research Laboratory                             fax: (303) 497-7259 <tel:%28303%29%20497-7259>
>> 325 Broadway R/GSD6
>> Boulder, CO 80303
>> 
>> 
>> 
>> 
>> 
>> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20180725/88c6221e/attachment.html>


More information about the slurm-users mailing list