One thing you could consider doing is setting a higher weight on the the
long nodes (cluster[37-100] in your example). This would cause jobs
submitted to the batch partition to attempt to schedule on low weight nodes
first, then the higher weight nodes. So "long" would only get used if a
job requ
https://github.com/SchedMD/slurm/blob/master/src/slurmctld/read_config.c
Line 2511 -- if the node has been scheduled exclusively, this field is set to
the uid of the user whose job(s) occupy the node.
> On Aug 8, 2018, at 18:14 , Ryan Novosielski wrote:
>
> Does anyone have any idea or a poi
I have an application group that would improve throughput if we could configure
jobs to run two on a node (but starting/finishing at individual job times)
packed by the scheduler rather than spread out and overlapped only when the
partition is fully loaded with one job per node. The users' workf
Does anyone have any idea or a pointer to documentation about what the node
“owner” field is in “scontrol show node ” like the below (set out by
*s):
[root@hal0099 ~]# scontrol show node hal0097
NodeName=hal0097 Arch=x86_64 CoresPerSocket=16
CPUAlloc=0 CPUErr=0 CPUTot=32 CPULoad=0.01
Avail
Ole Holm Nielsen writes:
> Bjørn, that is a different task.
I know, but related. Just meant as a tip for people who already use
pdsh.
--
Regards,
Bjørn-Helge Mevik, dr. scient,
Department for Research Computing, University of Oslo
signature.asc
Description: PGP signature
