SLURM Overview (Puma)
The new HPC system called Puma will have SLURM as a job scheduler rather than PBS Pro. SLURM has several advantages:
- More robust support for larger number of jobs in queue
- Used by national HPC groups (XSEDE and TACC) so it'll be easier to scale out to those systems.
PBS → SLURM Rosetta Stone
We've installed a software that converts some basic PBS Pro commands into SLURM commands automatically.
PBS → SLURM Commands
|PBS Command||SLURM Command||What the command does|
|qsub (for batch)||sbatch||Batch submission of jobs. Used for submitting many jobs that run without user input.|
|qsub -I (for interactive)||srun||A request for an interactive job where the user can input commands on the HPC machine.|
|qstat||squeue||Check the status of the jobs. Output all the jobs on the cluster.|
|qstat -u NetID||squeue -u NetID|
Check the status of the jobs for a specific user.
|qdel||scancel||Delete a job from the queue. Takes job ID as an argument.|
|qstat -Q||sinfo||View information about nodes and queues.|
PBS → SLURM Environment Variables
|PBS Env Variable||SLURM Env Variable|
Commands that are the same
PBS → SLURM Terminology
|PBS terminology||SLURM terminology||Definition|
What are available
How to see them