[slurm-users] clarification on slurm scheduler and the "nice" parameter

Matteo F mfasco984 at gmail.com
Wed Apr 15 06:55:38 UTC 2020


Hi,
just wanted to update this thread in case someone has a similar problem.

I managed to achieve this result by:
- changing SelectTypeParameters=CR_Core_Memory
To SelectTypeParameters=CR_CPU_Memory
- giving the low priority job a very high "nice" value, near the end of the
scale

With this configuration, when I submit an high priority job, slurm stop
running other low priority already-queued jobs to run my new job.

Thanks everyone
Matteo

On Tue, 14 Apr 2020 at 15:57, mercan <ahmet.mercan at uhem.itu.edu.tr> wrote:

> Hi;
>
> Did you restart slurmctld after changing
> "PriorityType=priority/multifactor"?
>
> Also your nice values are too small. It is not unix nice. Its range is
> +/-2147483645, and it race with other priority factors at priority
> factor formula. Look priority factor formula at
> https://slurm.schedmd.com/priority_multifactor.html
>
> Because of that, I think you should test with very high numbers for low
> priority nice such as --nice 100000.
>
> Regards;
>
> Ahmet M.
>
>
> 14.04.2020 16:18 tarihinde Matteo F yazdı:
> > Hello Lyn, thanks for your reply.
> > I checked my configuration; the PriorityType was set to
> > "PriorityType=priority/basic" initially, so my tests refer to that
> > configuration.
> > After you post, I set it to "PriorityType=priority/multifactor" and
> > ran the tests again: the results are the same.
> > I was hoping to find a quick fix but it seems that I need to dig into
> > the docs and play with the "factors".
> > Matteo
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20200415/99f248a4/attachment.htm>


More information about the slurm-users mailing list