I observed this same behavior with REQUEUE and CANCEL.  See
https://groups.google.com/d/msg/slurm-devel/xr3fcXKKmlg/B1keUh2_DAAJ

We currently use PreemptMode=SUSPEND,GANG PreemptType=preempt/partition_prio
in production, which correctly only suspends 1 job if a node is running
multiple jobs and only 1 needs to be preempted.

- 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 Sun, Jan 24, 2016 at 2:49 PM, alvaro gamboa <[email protected]>
wrote:

> Hi,
>
> I'm using preemption with REQUEUE, PARTITION-PRIO and job_requeue=1
> options.
>
> I have a node with 4 processors. When the 4 are working in different jobs with
> the same priority, and I execute another job with higuer priority, the 4
> jobs are reset and one of them is requeued. Why are they reset? The
> behavior should be that one of them is requeued and the other three continue
> with their work, right? How can I get this behavior?
>
> Thank you all, best regards
>

Reply via email to