Training hint in user/submitting-jobs uses /tmp/$USER — per-node on pcluster

Date: 2026-04-13. Impact: training.

← back to the Exercise Validation summary

What happened

user/submitting-jobs.md exercise 5 teaches --chdir by creating /tmp/$USER/test_chdir on the submitter host, then verifying the job ran there. On pcluster this can silently appear to work (the job "ran" and exits 0) but the file the training tells learners to cat afterwards lives on a different node's /tmp and isn't visible from the submitter.

Same root cause as lesson #1 — /tmp is per-node on elastic HPC. The exercise's teaching moment (how --chdir works) is unaffected by the filesystem choice; only the verification path needs to use a shared filesystem like $HOME or /shared.

Future content pass: either swap the path to $HOME/test_chdir, or add a deployment callout reminding learners that cross-node verification needs a shared mount.

Fix commits

• slurm-training-validation@7169aa2 — submitting-jobs.bats ex5 uses $HOME instead of /tmp