Rivanna HPC — Job Submission Guide

1. Web-Based Access (Open OnDemand)

Open OnDemand provides browser-based access to Rivanna. No VPN is required. The following walkthrough demonstrates how to launch a JupyterLab session with GPU resources.

Once a session is allocated, it continues running on the cluster independently of your local machine. You can safely close the browser and return to it later.

Step 1

Navigate to the UVA HPC login page and click “Launch Open OnDemand,” or go directly to ood.hpc.virginia.edu.

Step 2

You will land on the Open OnDemand dashboard. Click My Interactive Sessions in the top navigation bar.

Step 3

Select JupyterLab under Servers.

Other available apps include:

• Code Server — VS Code in the browser
• RStudio Server — for R workflows
• MATLAB Desktop / Web — for MATLAB users
• Desktop — full Linux desktop environment

JupyterLab is the most common choice for Python-based work.

Step 4

Configure your session:

• Rivanna/Afton Partition — select “GPU” for GPU-enabled jobs, or “Standard” for CPU-only jobs
• Number of hours — maximum wall time for your session (up to 72 hours)
• Number of cores — CPU cores allocated
• Memory — RAM in GB
• Work Directory — where your session’s files are stored (see table below)
• Allocation — your group’s allocation
• GPU type — choose a specific GPU or leave as “default”
• Number of GPUs — how many GPUs to request

The screenshot shows one example configuration. Adjust values to match your workload.

Click Launch when ready.

Available GPU types: a6000, a40, v100, a100, h200. If unsure, leave GPU type as “default” and the scheduler will assign any available GPU.

Work Directory Reference

Step 5

Once the session is running, click Connect to Jupyter to open JupyterLab in a new tab.

2. SSH Login

To connect to Rivanna, open a terminal on your local computer and run:

$ ssh -Y <computing_id>@login.hpc.virginia.edu

The -Y flag enables X11 forwarding for GUI applications.

VS Code (Recommended)

Most users connect through VS Code with the Remote – SSH extension, which provides a full editor, file browser, and integrated terminal on the cluster.

1. Install the Remote – SSH extension in VS Code.
2. Open the Command Palette (Ctrl+Shift+P) → Remote-SSH: Connect to Host.
3. Enter: <computing_id>@login.hpc.virginia.edu

3. Submitting Jobs

Rivanna is a shared computing cluster. To use its resources (GPUs, CPUs, memory), you must submit a job request to the scheduler (SLURM), which allocates resources on your behalf. There are two types of jobs:

3.1 Interactive Jobs

Request an interactive session with one A6000 GPU:

$ ijob -A mylab -p gpu --gres=gpu:a6000:1 -c 8 --mem=64G -t 4:00:00

To end the session:

$ exit

3.2 Batch Jobs

Create a SLURM script (e.g. job.slurm):

#!/bin/bash
#SBATCH -A mylab
#SBATCH -p gpu
#SBATCH --gres=gpu:a6000:1
#SBATCH -c 8
#SBATCH --mem=64G
#SBATCH -t 8:00:00
#SBATCH -o output_%j.log
#SBATCH -J my_job

module load cuda python
python my_script.py

Submit the script:

$ sbatch job.slurm

4. Monitoring & Managing Jobs

5. Common Pitfalls

6. References

• Rivanna HPC Login — UVA Research Computing
• SLURM Job Manager — UVA Research Computing
• UVA Anywhere VPN Setup
• VS Code Remote – SSH Extension