Testing & running your first job

Overview

Questions

• How do you verify that a Slurm cluster is working correctly?
• How do you submit and monitor batch jobs?

Objectives

• Check cluster health using sinfo
• Submit a batch job with sbatch and monitor it with squeue
• Run an interactive job using srun
• Verify the shared filesystem works across nodes

Before submitting any work, verify that the cluster is healthy and all nodes are visible to the scheduler.

Check cluster status

---

From the login node, run:

BASH

sinfo

You should see your compute node listed as idle:

PARTITION        AVAIL  TIMELIMIT  NODES  STATE NODELIST
pixiecluster*       up   infinite      1   idle pixie02

If the node shows down or unknown, check that slurmd is running on the compute node:

BASH

ssh pixie02 systemctl status slurmd

Submit a minimal batch job

---

Create a file called hello.sh:

BASH

#!/bin/bash
#SBATCH --job-name=hello
#SBATCH --output=hello-%j.out
#SBATCH --ntasks=1

echo "Hello from $(hostname) at $(date)"

Submit it:

BASH

sbatch hello.sh

Slurm will print a job ID, e.g. Submitted batch job 1.

Check job status

---

BASH

squeue

While the job is running you will see it listed. Once it completes the queue will be empty. Check the output file (replace 1 with your job ID):

BASH

cat hello-1.out

Expected output:

Hello from pixie02 at Fri Jun 13 10:00:00 BST 2026

The hostname should be your compute node, not the login node.

Run an interactive job

---

For debugging it is often useful to get a shell directly on a compute node:

BASH

srun --pty bash

Your prompt will change to reflect the compute node hostname. Type exit to return to the login node.

Test the shared filesystem

---

Jobs can read and write to /sharedfs from any node. Verify this round-trips correctly:

BASH

#!/bin/bash
#SBATCH --job-name=sharedfs-test
#SBATCH --output=sharedfs-%j.out
#SBATCH --ntasks=1

echo "written by $(hostname)" > /sharedfs/test.txt
cat /sharedfs/test.txt

After the job completes, confirm the file is visible from the login node:

BASH

cat /sharedfs/test.txt

Test a multi-node job

---

If you have more than one compute node, verify that Slurm can span them:

BASH

#!/bin/bash
#SBATCH --job-name=multinode
#SBATCH --output=multinode-%j.out
#SBATCH --ntasks-per-node=1
#SBATCH --nodes=2

srun hostname

The output file should contain one line per node.

Key Points

• Use sinfo to check that all nodes are visible and in idle state before submitting jobs
• sbatch submits batch jobs; squeue monitors the queue; job output goes to the file specified by --output
• srun --pty bash opens an interactive shell on a compute node for debugging