As the cluster is a shared piece of equipment, please adhere to the following rules.
- No-blame culture. If you stumble across a problem, report it with falke regardless of cause.
- Cancel jobs that are useless/not producing valuable results ASAP
- Make space for others!
- Occupying the entire cluster for an hour/over night is ok
- Do not allocate the entire cluster for weeks
- Offload large datasets from your home
~to your own machine when not needed - If you have multi-node, long-running jobs, clear them with falke first.
- Other people may have tighter deadlines than you
- Keep an eye on your jobs (daily)
- Your script may have failed, leading to wasted cpu-time for others
- Use the
--mail-type=FAILdirective to get emails on failing jobs
- Clean up after yourself in
/scratch