HPCE/slurm-efficiency-tools
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

improve docstring

Browse Source
This commit is contained in:
feichtinger
2026-07-01 11:17:56 +02:00
parent 7437e5f4d0
commit 4b397a2835
1 changed files with 12 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
slurm-eff-tool.py
+12 -9
View File
@@ -1,21 +1,24 @@
#!/usr/bin/env python3
"""
slurm-eff-tool.py - Slurm job efficiency reporting and investigation tool.
slurm-eff-tool.py - Slurm job efficiency analysis and reporting and tool.
Also targeted at automation and backfeeding use cases.
Efficiency definitions:
CPU_Eff = TotalCPU_seconds / CPUTimeRAW * 100
Time_Eff = ElapsedRaw_seconds / (TimelimitRaw_minutes * 60) * 100
CPU_Eff = CPU_time / walltime * 100
Time_Eff = walltime / timelimit * 100
Mem_Eff = 100 (slurm recorded mem usage) / (mem allocated by system)
This script intentionally asks sacct for raw parsable fields and caches those rows.
TODO: allow to cache a faster already parsed and binary format. Keep sacct text option for
debug option
Waste definitions:
waste_CPU = walltime * (100 - CPU_Eff) * n_cpus [hours]
waste_Mem = walltime * (100 - Mem_eff) * \
(req_mem - (n_cpus * default_mem_per_cpu )) [GB*hours]
waste_total = waste_CPU + waste_Mem/default_mem_per_cpu
2006 D. Feichtinger <derek.feichtinger@psi.ch>
"""
# initial version of script produced by vibe coding of an exact functionality
# description using chatgpt-5.5
# 2026 D. Feichtinger <derek.feichtinger@psi.ch>
from __future__ import annotations