Slurm job debug template
Skill ma-compbio-lab/SkillFoundry/skills/hpc/slurm-job-debug-template
Use this skill to render, submit, and inspect a minimal Slurm smoke job before submitting heavier workloads on a real cluster.From its SKILL.md
npx -y skills add ma-compbio-lab/SkillFoundry --skill slurm-job-debug-templateAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
SKILL.md
2.3 KB, 558 tokens by cl100k_base, as published. Nobody here has run it
Purpose
Generate a conservative sbatch template, submit it to Slurm, and capture the accounting record needed to verify the cluster path end to end.
When to use
- You need a tiny Slurm smoke job.
- You want a reusable starting point for queue, environment, and log validation.
When not to use
- You need multi-node or GPU tuning guidance beyond a first smoke job.
Inputs
- Command string
- Optional job name, partition, runtime, memory, and output path
Outputs
- Rendered
sbatchscript - JSON submission and
sacctsummary
Requirements
- Python 3.13+
sbatch,squeue, andsacct- A real Slurm cluster
Procedure
- Run
python3 skills/hpc/slurm-job-debug-template/scripts/render_sbatch.py --command "echo hello" --job-name smoke. - Inspect the generated script or save it with
--out. - Submit a verified smoke job with
python3 skills/hpc/slurm-job-debug-template/scripts/submit_smoke_job.py --partition cpu --job-name slurm-smoke --sleep 2 --out slurm/reports/slurm-smoke.json. - Inspect the returned
job_id, the log paths inslurm/logs/, and theaccountingblock fromsacct.
Validation
- Renderer exits successfully.
- Output contains
#SBATCHdirectives and the command body. - Submitted smoke job reaches
State=COMPLETED. ExitCodeis0:0.
Failure modes and fixes
- Missing partition/account information: add them before submission.
- Output paths wrong: switch to cluster-appropriate scratch or log paths.
sacctlags briefly after completion: retry once accounting catches up.
Safety and limits
- Keep resource requests small for initial smoke jobs.
- Use a CPU partition and a short walltime for smoke checks.
Examples
python3 .../render_sbatch.py --command "python analysis.py" --partition short --time 00:10:00 --mem 2Gpython3 .../submit_smoke_job.py --partition cpu --job-name slurm-smoke --sleep 1
Provenance
- Slurm quick start: https://slurm.schedmd.com/quickstart.html
- Slurm
sbatchreference: https://slurm.schedmd.com/sbatch.html sacctreference: https://slurm.schedmd.com/sacct.html
Related skills
snakemake-toy-workflow-starter
What ships with it: 10 files
9.2 KB alongside SKILL.md, 2 of them executable
assets/
- README.md88 B
- slurm_smoke_report.json687 B
- smoke_job.sbatch281 B
examples/
- README.md210 B
- smoke_job.sbatch281 B
scripts/
- render_sbatch.pyruns1.9 KB
- submit_smoke_job.pyruns4.3 KB
tests/
- README.md282 B
- metadata.yaml918 B
- refs.md218 B