Job detail

Press Enter on a job row to open the detail screen. It runs scontrol show job <id> for static info, sstat for live memory usage, and (for GPU jobs) srun --jobid=<id> --overlap nvidia-smi … for per-GPU utilisation.

Job detail screen

How to read it

  • Time / Memory bars are computed as RunTime / TimeLimit and MaxRSS / ReqTRES.mem. Bars over 80 % are red; 50 – 80 % yellow; below 50 % green.

  • GPU utilisation comes from nvidia-smi --query-gpu=… issued via srun --overlap so the readings come from inside the job’s allocation. If your cluster restricts srun --overlap, the per-GPU bars stay blank — the Allocated: 4x l40s line still appears from scontrol.

  • Log paths are read directly from scontrol’s StdOut / StdErr fields, so they reflect Slurm’s actual file path including any %j / %a expansion already resolved.

Keybindings

Key

Action

o

Open the log viewer on StdOut.

e

Open the log viewer on StdErr.

v

Open the batch script viewer (the submitted sbatch script).

c

scancel this job — see Cancelling jobs.

y

Copy a value to the clipboard. Cycles: job ID → stdout path → stderr path → work dir.

j / k / g / G

Scroll within the detail body.

Esc / q

Back to the job list.