Appendix B — Slurm Basics

This appendix summarizes the essential Slurm commands required for CADRE Year‑2 training. These commands apply to Hercules and most HPC systems.

B.1 Submitting Batch Jobs

Submit a job script:

sbatch job.slurm

A minimal example job script:

#!/bin/bash
#SBATCH --job-name=test
#SBATCH --qos=normal
#SBATCH --time=01:00:00
#SBATCH --ntasks=1

srun hostname

B.2 Interactive Jobs

Request an interactive compute session:

salloc --qos=<qos> --time=01:00:00 --ntasks=1

Then run commands inside the allocated node.

Cancel an interactive session:

exit

B.3 Monitoring Jobs

List your jobs:

squeue -u $USER

List all jobs:

squeue

Show job details:

scontrol show job <jobid>

B.4 Job Output and Errors

Slurm writes output to files such as:

slurm-<jobid>.out

Check job errors inside this file if a job fails.

B.5 Useful Slurm Commands

• Check partitions/QOS:

  sinfo
  

• Check job efficiency:

  seff <jobid>
  

• Cancel a job:

  scancel <jobid>
  

B.6 Summary

This appendix provides the minimal Slurm commands needed to submit, monitor, and debug jobs for CADRE Year‑2 training workflows.