From zero to your first job in ~30 minutes
SeaWulf is available to the campus community and external collaborators, including Brookhaven National Laboratory personnel and New York State companies and agencies. Access is project-based: projects are established by Stony Brook faculty and staff, and individual accounts are authorized under them โ see requesting accounts.
1 ยท Connect
Open OnDemand (recommended for beginners) gives a browser GUI โ see the overview. Or connect by SSH:
# For general computing (Rocky Linux 9.6) ssh -X SBU_NETID@milan.seawulf.stonybrook.edu # For legacy applications (CentOS 7) ssh -X SBU_NETID@login.seawulf.stonybrook.edu
Both require DUO authentication. Windows users: consider MobaXterm.
2 ยท Know the architecture
- login.seawulf.stonybrook.edu (login1/login2) โ older Intel Haswell nodes and legacy GPU nodes
- milan.seawulf.stonybrook.edu (milan1/milan2) โ Intel Skylake (40-core), AMD EPYC Milan (96-core), and Sapphire Rapids (HBM) nodes for most workloads
Critical: Never run intensive computations on login nodes โ they're shared entry points for setup and submission only (see login node etiquette).
3 ยท Load software with modules
module avail module load slurm module load gcc/12.3.0 module list
4 ยท Your first job
- Prepare a workspace.
mkdir ~/my_first_job && cd ~/my_first_job module load slurm
- Create
hello.slurm.#!/bin/bash #SBATCH --job-name=hello_world #SBATCH --partition=short-40core-shared #SBATCH --nodes=1 #SBATCH --ntasks=1 #SBATCH --mem=1G #SBATCH --time=00:05:00 #SBATCH --output=hello_%j.out #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)"
- Submit it.
sbatch hello.slurm
Returns something like Submitted batch job 12345.
- Monitor and view output.
squeue -u $USER cat hello_12345.out
5 ยท Job states
| State | Description |
|---|---|
| PD (Pending) | Waiting for resources |
| R (Running) | Currently executing |
| CG (Completing) | Job is in the process of completing |
Common issues
- Job won't start โ confirm partition requirements, request fewer resources, or validate with
sbatch --test-only script.slurm - Connection problems โ repeated failed attempts can trigger IP or DUO blocking; open a support ticket if it persists
- GUI won't launch โ add the
-Xflag (see X11 forwarding)
Next: the job-script guide, the queue tables, and the cheat sheet.
Applies to