This is much less complex in 14.11. Here is what I use to set qos to
partition name if qos isn't explicitly set (this is the entire script):
function slurm_job_submit(job_desc, part_list, submit_uid)
-- ensure a partition is set
if job_desc.partition == nil then
job_desc.partition = "makena"
end
if job_desc.qos == nil or job_desc.qos == "normal" then
job_desc.qos = job_desc.partition
end
return slurm.SUCCESS
end
function slurm_job_modify(job_desc, job_rec, part_list, modify_uid)
return slurm.SUCCESS
end
slurm.log_info("initialized")
return slurm.SUCCESS
On Fri, Feb 20, 2015 at 01:36:46PM -0800, Trey Dockendorf wrote:
> We took the method of using a job_submit plugin and per-partition QOS (with
> some sharing). See https://gist.github.com/treydock/b964c5599fd057b0aa6a.
> Doesn't work with 14.11, only 14.03 (at least that's what we've tested on).
>
> - Trey
>
> =============================
>
> Trey Dockendorf
> Systems Analyst I
> Texas A&M University
> Academy for Advanced Telecommunications and Learning Technologies
> Phone: (979)458-2396
> Email: [email protected]
> Jabber: [email protected]
>
> On Fri, Feb 20, 2015 at 3:15 PM, Chris Read <[email protected]> wrote:
>
> > Thanks guys - I think we'll wait for 15.08
> >
> > Chris
> >
> > On Fri, Feb 20, 2015 at 2:40 PM, Schmidtmann, Carl <
> > [email protected]> wrote:
> >
> >>
> >> The simplest solution currently available is to write a job submit script
> >> to set the QOS based on the partition name. This means a single QOS per
> >> partition and requires a QOS be created for each partition. Or a very
> >> complicated job submit script.
> >>
> >> But the previous message with this subject said that is changing in an
> >> upcoming release and available now for testing in a pre-release version.
> >>
> >> Carl
> >>
> >>
> >> > On Feb 19, 2015, at 3:01 PM, Chris Read <[email protected]> wrote:
> >> >
> >> > Greetings...
> >> >
> >> > We currently rely heavily on QOS based time and resource limits, but up
> >> until recently have been running in a single partition. Now that we have
> >> multiple partitions we've seen the need for some of these limits (for
> >> example GrpCPUs) to be limited per partition.
> >> >
> >> > I see no easy way to do that right now other than creating new
> >> partitions, which I don't want to do. Are there any plans to make this easy
> >> to do?
> >> >
> >> > Anyone else interested in such a feature?
> >> >
> >> > Thanks,
> >> >
> >> > Chris
> >> >
> >>
> >
> >
--
andy wettstein
hpc system administrator
research computing center
university of chicago
773.702.1104
