The script calls sacct --json and parses it.
It then outputs a tab-separated table with the requested cpu, memory and time and their actual usage.
This helps to tune your request size and get better scheduling.
For example, when available cputime is much higher than actual cpu time, your job has not used all provided cores at 100% and could probably do with less.
usage: sacct-analyze.py [-h] [--name NAME]
optional arguments:
-h, --help show this help message and exit
--name NAME Should have these contents in the name
python3 sacct-analyze.py > results.tsv
| Name | Requested CPU | Available CPU time (hours) | Used CPU time (hours) | CPU Efficiency (%) | Requested Memory (GiB) | Used Memory (GiB) | Memory efficiency (%) | Requested Time (hours) | Used Time (hours) | Time Efficiency (%) |
|---|---|---|---|---|---|---|---|---|---|---|
| call-deepVariantTask-2 | 4 | 17.83 | 17.13 | 96.03% | 48.00 | 8.03 | 16.73% | 83.33 | 4.46 | 5.35% |
| call-deepVariantTask-6 | 4 | 17.30 | 16.45 | 95.09% | 48.00 | 7.98 | 16.63% | 83.33 | 4.33 | 5.19% |
| call-deepVariantTask-1 | 4 | 50.05 | 48.72 | 97.33% | 48.00 | 9.30 | 19.38% | 83.33 | 12.51 | 15.02% |
| call-deepVariantTask-3 | 4 | 50.53 | 49.26 | 97.49% | 48.00 | 9.39 | 19.57% | 83.33 | 12.63 | 15.16% |
| call-deepVariantTask-0 | 4 | 40.17 | 39.08 | 97.29% | 48.00 | 8.04 | 16.74% | 83.33 | 10.04 | 12.05% |
| call-deepVariantTask-4 | 4 | 35.69 | 34.60 | 96.94% | 48.00 | 8.06 | 16.79% | 83.33 | 8.92 | 10.71% |
In this example all the four provided cores are kept busy during the job runtime. The amount of memory and time needed was much less than initially anticipated however.