Getting an overview of available Slurm partitions and their current job
load is a non-trivial task.
The great "spart" tool described as "A user-oriented partition info
command for slurm" (https://github.com/mercanca/spart) written by Ahmet
Mercan solves this problem. The "spart" tool is written in C and linked
to Slurm libraries.
I have now written a new "showpartitions" tool
(https://github.com/OleHolmNielsen/Slurm_tools/tree/master/partitions)
which has the same features as "spart". It should be useful to end
users as well as system administrators.
The "showpartitions" tool is a shell/awk script which parses the outputs
from sinfo and squeue Slurm commands to obtain the partitions overview.
The columns are ordered a little differently from "spart", and color
coding helps locating under- or over-utilized partitions easily.
The tool has been tested at a handful of Slurm sites. I'd be really
interested in comments and suggestions from other Slurm sites. In
particular, it would be great if someone with access to federated
clusters could verify the output.
In addition, the small "showhidden" tool prints a Slurm cluster
partition status for hidden and root-only partitions, if your site uses
these features. This information is not easily accessible using the
sinfo command, since no "sinfo --only-hidden" option exists.
Acknowledgments:
Ahmet Mercan <ahmet.mer...@uhem.itu.edu.tr> has been really supportive
in testing showpartitions during development and comparing the output to
spart.
Thorsten Deilmann from University of Wuppertal has offered a number of
useful suggestions, including the colored output.
Best regards,
Ole
--
Ole Holm Nielsen
PhD, Senior HPC Officer
Department of Physics, Technical University of Denmark
