SLURM multiple cores per node

Asked by Joshua Isaacson on 2018-09-13

I am trying to run some code on a SLURM cluster that has 12 cpus per node. However, with the following options, I am unable to get the code to run on more than a single cpu per node.

cluster_type= slurm

and I went into the file bin/internal/cluster.py, and added
--ntasks-per-node=12

to the submission script. I believe that Madgraph is only submitting it with one job at a time, so where in the code will I have to modify Madgraph to allow it to submit 12 jobs to a single node?

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Joshua Isaacson
Solved:
2018-09-13
Last query:
2018-09-13
Last reply:
2018-09-13

Hi,

I do not see the point of doing that...
SLURM should decide where to assign your jobs.
If a node is free and has 12 slot, then SLURM will assign 12 of the slot to your job.
If only 6 are available then 6 of the madgraph jobs will be allowed to run.

If you want to be the only user running on a single node, you might use the --exclusive=user argument

Cheers,

Olivier

> On 13 Sep 2018, at 16:32, Joshua Isaacson <email address hidden> wrote:
>
> New question #673797 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/673797
>
> I am trying to run some code on a SLURM cluster that has 12 cpus per node. However, with the following options, I am unable to get the code to run on more than a single cpu per node.
>
> cluster_type= slurm
>
> and I went into the file bin/internal/cluster.py, and added
> --ntasks-per-node=12
>
> to the submission script. I believe that Madgraph is only submitting it with one job at a time, so where in the code will I have to modify Madgraph to allow it to submit 12 jobs to a single node?
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Joshua Isaacson (isaacs21) said : #2

Hi Olivier,

Ok. I guess I was just misinterpreting the output from Madgraph for how the jobs are being run.

Thank you!

Regards,
Josh