[slurm-users] Running two multiprocessing jobs in one sbatch

[Brian Andrus]

Is there a reason to run them as a single job?

It may be easier to just have 2 separate jobs of 16 cores each.

If there are dependency requirements, that is addressed by adding any 
dependencies to the job submission.

Brian Andrus

On 7/25/2020 2:50 AM, Даниил Вахрамеев wrote:
> Hi everyone!
>
> I have SLURM cluster with several nodes with 16 vcpus per node. I've 
> tried to run the following code:
>
> |#SBATCH --nodes 2 #SBATCH --ntasks 2 #SBATCH -c 16 srun --exclusive 
> --nodes=1 program1 & srun --exclusive --nodes=1 program2 & wait |
>
> |program1| and |program2| needs 16cpus each and I expected that 2 
> nodes with 32 cores would be allocated and |program1| would be ran on 
> the first node and |program2| on the second one, but I got the 
> following error message:
>
> |srun: error: Unable to create step for job 364966: Requested node 
> configuration is not available |
>
> If I use only |--nodes| and |--ntasks| keys, sbatch allocates 2 nodes 
> with 2 cpus and if I use |--nodes| and |-c| options, I get message 
> that |--ntasks| should be defined.
>
> If I set |--ntasks=1|, SLURM set nnodes to 1.
>
> How can I run this two programs in one batch, each on one node and 16 
> vcpus?
>
> ------
>
> Kind regards,
>
> Daniil Vakhrameev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20200725/f56d96bb/attachment.htm>