Jobs¶
See the jobs on a cluster, open their output files, and read why a job is waiting. You can also submit a Slurm script, cancel a job, and get a notice when a job ends.
See your jobs¶
Jobs lists work on one cluster. Pick the cluster, then choose whose jobs to show.
- Open Jobs.
- Pick the cluster at the top.
- Under Which jobs, choose a view.
- My jobs
- Jobs that are running or waiting now.
- My group
- Jobs from everyone in your groups.
- My history (7 days)
- Your jobs from the last week, including finished ones.
- All jobs
- Everyone's jobs on the cluster.
State filters the list by running, waiting, finished, and similar states. The search box finds a job by id, name, person, or queue. For example, search for jdoe.
Select a job to open its details. You see the time used, the folders, and the output files. Each folder and each output file has Copy path. All fields shows everything Slurm knows about the job.
Submit a job¶
You submit a job from a Slurm script open in Code.
- Open the script in Code from Files, from a project, or from a job's folder.
- Select Submit job in the bar above the script.
The new job appears in Jobs.
Why a job waits¶
When a job has not started, Jobs writes the reason in plain words. For example:
- Waiting: higher-priority jobs go first
- Waiting: not enough free nodes yet
- Waiting: maintenance has to finish first
- Waiting: your group is out of SUs
- Won't start: longer than this queue allows
Cancel a job¶
Cancel a job to stop it. HPC Hub asks you to confirm first.
Warning
Cancelling a job stops everything running in it. When the job is a node HPC Hub got for you, the action is Release node. Your terminals on that node stop too.
- Select ⋮ on the job's row.
- Select Cancel job.
- Confirm when HPC Hub asks.
To cancel several jobs, tick them and use the bar that appears.
Notifications¶
While a cluster is connected, HPC Hub checks your jobs about once a minute and tells you when one ends. Select the notification to open Jobs.