My job is queued up or denied by Slurm

My job is queued up (“PD – Pending” in Slurm)

To know the start time (at the latest) of your job:

squeue --start

Your job may start earlier as running jobs often end before their walltime.

If the Slurm reason is:

MaxCpuPerAccount: each job queue has specific constraints which are defined on this page. If the resources requested by your job would exceed this limit, your job is put on queue until other jobs from your project terminate.

Resources: this means there are not enough resources currently available on the cluster to meet your request in terms of resources and/or walltime.

Priority: there are other jobs pending with a higher priority than your job. Read more about priority for pending jobs.

Dependency: your job depends on an event before it can start. Typically, it has to wait for another job to finish. For instance, if you used the dependency after Slurm option (view details here) in your job definition. Or your job may require a license and all the available licenses are currently in use.

My job is denied by Slurm

Many reasons can cause your job is denied by Slurm, including:

  • You didn’t mention your project name with the option “account” or -A.
  • The resources/walltime requested exceed the limits set for the desired queue. View the limits here.
  • You want to use the smp or gpu queue whereas this access has not been granted to your project (the access is only granted upon request).
  • You’re not a permanent researcher and the end date of your contract as mentioned in your user charter has been reached. (You will receive reminders one month before the expiry and your supervisor must contact us if your contract has been extended.)
  • Your project hasn’t been renewed. The project’s leader receives several reminders each year asking to fill this form. Access to the compute nodes is automatically suspended if the form isn’t filled each year (error: QOSGrpCPUMinutesLimit).