Quick Start Guide

This guide takes you from a brand-new account to your first finished job on SeaWulf in about 30 minutes. You'll connect, load software, write a tiny job script, submit it, and read the result.

On this page: Before you start · 1. Connect · 2. Load software · 3. Your first job · 4. Watch it run · What next · If something goes wrong

Before you start

• A SeaWulf account. Accounts sit under a project led by a Stony Brook faculty member. See Requesting accounts & project numbers.
• Duo set up on your phone. Every login asks for it. See Duo authentication.

1. Connect

The easiest start is Open OnDemand in your browser at sn-ood.seawulf.stonybrook.edu. Sign in with your NetID, approve Duo, then open the SeaWulf Cluster Shell Access app for a terminal. See the OnDemand overview.

Prefer a terminal on your own computer? Use SSH:

ssh <netid>@milan.seawulf.stonybrook.edu

Windows users can use MobaXterm or the built-in Windows Terminal. Connecting to the cluster has step-by-step instructions and explains which login address to use.

2. Load software

Software on SeaWulf comes as modules that you load when you need them. Start by loading Slurm, the job scheduler, which you'll need in every session:

module load slurm             # job commands: sbatch, squeue, scancel
module avail                  # see what software is available
module load gcc/12.3.0        # load a specific package and version
module list                   # check what's loaded

More in Using modules and the software catalog.

3. Your first job

1. Make a folder for it.

   mkdir ~/my_first_job && cd ~/my_first_job

2. Create a job script called hello.slurm. Open an editor with nano hello.slurm, paste this in, then save with Ctrl+O and exit with Ctrl+X:

   #!/bin/bash
   #SBATCH --job-name=hello_world
   #SBATCH --partition=short-40core-shared   # a small, shared queue
   #SBATCH --nodes=1
   #SBATCH --ntasks=1
   #SBATCH --mem=1G
   #SBATCH --time=00:05:00                   # 5 minutes is plenty
   #SBATCH --output=hello_%j.out             # %j becomes the job ID
   #SBATCH --error=hello_%j.err
   
   echo "Hello from SeaWulf!"
   echo "Job ID: $SLURM_JOB_ID"
   echo "Running on node: $(hostname)"
   echo "Date and time: $(date)"

3. Submit it.

   sbatch hello.slurm

   Slurm replies with something like Submitted batch job 12345. That number is your job ID.
4. Read the result once it finishes (use your own job ID):

   cat hello_12345.out

4. Watch it run

squeue -u $USER

The ST column shows where your job is:

When the job disappears from squeue, it's done. Check how it went with sacct -j <jobid>.

What next

• Writing job scripts: the full template and every common option.
• Queue selection guide: pick the right queue for your job size and length.
• Interactive jobs: a live shell on a compute node for testing.
• File transfer: get your data onto the cluster.
• Job script builder: answer a few questions and get a script.
• HPC cheat sheet: the commands you'll use every day, on one page.

If something goes wrong

sbatch --test-only hello.slurm

Still stuck? When to ask for help explains how to reach HPC support and virtual office hours.