Custom squeue

Custom squeue for a slurm server

Parses the output of scontrol show job -d and sinfo and shows the user-relevant information, featuring a GPU count of the cluster resources and per-user information.

When run with the --default settings it will show a breakdown of free and utilized resources (nodes), the pending, running and past jobs of the current user, an overview of the number of resources used by each other user and the incoming load on the cluster.

When run with the flag --jobs it will show all of the running + pending jobs on the cluster, similar to squeue.

It can also be used interactively as a python library, using the output of JobMaster to inspect all of the cluster jobs:

from job_master import JobMaster
jobs = JobMaster()

The nodes can be interacted with in a similar way:

from nodes import NodeMaster
nodes = NodeMaster()
print(nodes)

Example usage:

python custom_squeue_v2.py --default

Example entry in .bashrc

alias custom_squeue="watch -n 2 python <PATH_TO_FOLDER>/custom_squeue_v2.py --default"