| Command | PBS | SLURM |
|---|---|---|
| Queues / partitions | qstat -q |
sinfo --summarize |
| Task list | qstat -r, qstat -i |
squeue |
| User task list | qstat -u <username> |
squeue -u <username> |
| Submitting tasks | qsub |
sbatch / srun / salloc |
| Task status | qstat |
squeue <job_id>, scontrol show job <job_id> |
| Deleting a task | qdel <job_id> |
scancel <job_id> |
| Deleting all user tasks | qselect -u <username> | xargs qdel, qdel -u <username> |
scancel -u <username> |
| Interactive session in a pseudo-terminal | qsub -I |
srun --pty <program> |
Directives for the slurm queueing system can be included in a file that also contains directives for the PBS queueing system. The interpreter will treat directives that do not belong to it as comments and will not disrupt the task execution.
| Argument | #PBS |
#SLURM |
|---|---|---|
| Task name | -N <name> |
-J <name> |
| Queue / partition | -q <name> |
-p <name> |
| Number of nodes | -l select=<number> |
-N <number> |
| Total number of subtasks / MPI processes | N/A | -n <number> |
| Total number of computing cores per subtask | N/A | -c <number> |
| Number of computational cores | -l ncpus=<number> |
N/A |
| Number of MPI processes per node | -l mpiprocs=<number> |
--ntasks-per-node=<number> |
| Amount of RAM per node | -l mem=<amount> |
--mem=<amount> |
| Time limit | -l walltime=<hh:mm:ss> |
-t <minutes>, -t <days-hh:mm:ss> |
It is worth noting that allocating the number of cores in the SLURM queueing system defaults to total sizes, not per node, as in the PBS system.