Skip to content

Your first job

This page walks you through submitting a job to COSMOS for the first time — writing a script, submitting it, checking its progress, and reading the output. The whole process takes about five minutes.

1. Open a terminal on COSMOS

Log in via SSH or the HPC Desktop if you have not already. All commands below are run on COSMOS, not your local machine.

2. Write a job script

A job script is a plain text file that tells SLURM what resources your job needs and what commands to run. Create one with a text editor — nano is a good choice if you are new to the command line:

nano first_job.sh

Paste in the following:

#!/bin/bash
#SBATCH -J my_first_job
#SBATCH -t 00:05:00
#SBATCH --qos=test

echo "Job started at: $(date)"
echo "Running on node: $(hostname)"
echo "Hello from COSMOS!"
echo "Job finished at: $(date)"

Save and close the file. In nano: Ctrl+O then Enter to save, Ctrl+X to exit.

What each line does

Line Meaning
#!/bin/bash Use bash to run this script
#SBATCH -J my_first_job Give the job a name (visible in the queue)
#SBATCH -t 00:05:00 Request up to 5 minutes of compute time
#SBATCH --qos=test Use the test queue — jobs get elevated priority (max 1 hour, max 2 nodes)

Always use --qos=test while testing

The test queue gives your job elevated priority so you get results quickly. Remove it when you move on to production runs.

3. Submit the job

sbatch first_job.sh

SLURM will confirm with a job ID:

Submitted batch job 12345

Note the number — you will use it to find the output file.

4. Monitor the job

Check the status of your job with jobinfo:

jobinfo -u $USER

You will see something like this while the job is pending:

------------------------ Running jobs ------------------------
(none)

------------------------ Waiting jobs ------------------------
  JOBID PARTITION       NAME     USER  ST       TIME  NODES NODELIST(REASON)
  12345      lu48 my_first_  yourid  PD       0:00      1 (None)

And like this while it is running:

------------------------ Running jobs ------------------------
  JOBID PARTITION       NAME     USER  ST       TIME  NODES NODELIST(REASON)
  12345      lu48 my_first_  yourid   R       0:02      1 cn001

Once your job no longer appears in the list, it has finished. The key status codes are:

Code Meaning
PD Pending — waiting for resources to become available
R Running
CG Completing — cleaning up after the job

5. Read the output

SLURM writes all output to a file named slurm-<jobid>.out in the directory where you ran sbatch. For job 12345:

cat slurm-12345.out

You should see:

Job started at: Mon Mar 30 10:42:01 CET 2026
Running on node: cn001
Hello from COSMOS!
Job finished at: Mon Mar 30 10:42:01 CET 2026

The node name (cn001) confirms the job ran on a compute node, not the login node.

If the output file is empty or missing

Wait a few seconds and try again — SLURM may not have flushed the file yet. If the file contains an error message, see When your job fails for a guide to diagnosing common problems.

6. Using software in your job

Most research workflows require specific software. COSMOS provides applications through a module system — if you have not used it yet, read Using software modules first.

The key rule: modules loaded in your terminal are not available on the compute node. Load them explicitly in your job script:

#!/bin/bash
#SBATCH -J my_python_job
#SBATCH -t 00:05:00
#SBATCH --qos=test

module purge
module load GCCcore/13.2.0
module load Python/3.11.5-GCCcore-13.2.0

python my_script.py

Where to go next

Your next steps depend on what you want to run:

Author: (LUNARC)

Last Updated: 2026-03-30