Visualize instantaneous allocated resources (CPUs or GPUs) on Slurm clusters over time.
Queries sacct for job records, uses an event-based sweep algorithm to compute a
1-second-accurate allocation step function, and renders overlaid time-series in the
terminal via tplot.
Designed for HPC administrators and leadership to visualize fair-share usage patterns across accounts, users, and QOS levels.
uv tool install .
sacct-plot [-h] [-v] [-u USER] [-A ACCOUNT] [-r PARTITION] [-q QOS]
[-s STATE] [-S STARTTIME] [-E ENDTIME]
[--by {account,user,qos}] [--gpu] [--bucket INTERVAL]
[--sum | --mean | --max | --min] [--top N] [--stacked] [--data]
# Show CPU allocation over the past 7 days
sacct-plot -S 2026-02-11
# Overlay by account, show top 5
sacct-plot -S 2026-02-11 --by account --top 5
# GPU allocation, bucketed by hour
sacct-plot -S 2026-02-11 --gpu --bucket 1h
# Dump data instead of plotting
sacct-plot -S 2026-02-11 --by user --data
MIT — see LICENSE.