Search code examples
cluster-computingslurmsbatchsacct

How do I find the queuing time for completed SLURM jobs?

I am trying to find a way to see the waiting time of completed jobs until they started on a cluster with SLURM as the Workload manager.

I know the "sacct" command can print a lot of information about completed jobs but I don't find anything useful for my task.

Perfect would be either:

  • The time the job was queuing until it started
  • A timestamp or date of the time I submitted the job (then I can calculate back the waiting time, because "sacct" gives information about the time the job started with the "Start" field)

Do you have any useful tips?

Solution

  • The queue time for a job can be obtained from the sacct command like this:

    sacct -X  -j <jobid> -o Reserved

    The Reserved column is, (from the sacct manpage)

    Reserved How much wall clock time was used as reserved time for this job. This is derived from how long a job was waiting from eligible time to when it actually started. Format is the same as Elapsed.

    Note this has nothing to do with so-called "advanced resource reservations" in Slurm.

    Basically the time a job is "submitted" and that job is "eligible" is most often the same, except in some cases, for instance when the job is submitted with --hold, or with --dependency.

    UPADTE: recent versions of Slurm have this parameter named to Planned:

    How much wall clock time was used as planned time for this job. This is derived from how long a job was waiting from eligible time to when it started or was cancelled. Format is the same as Elapsed.