2 Getting Started
2.1 Account Setup
- Request access from your PI or supervisor
- Read and understand the Acceptable Use Policy
- Obtain your UnityID and ensure VPN access if working off-campus
2.2 Logging In
There are multiple ways to log onto Hazel:
Web browser (OnDemand): https://servood.hpc.ncsu.edu
- Select NC State University from the drop down menu
- Once logged on, use the tabs at the top to access files, submit jobs, use apps, etc.
To open up a terminal session, use the clusters tab and select >_ Hazel Cluster Shell Access
Using OnDemand may be especially useful if you have a windows machine, as terminal access may be challenging based on your setup.
Terminal (SSH):
$ ssh [UnityID]@login.hpc.ncsu.eduTo enable GUI applications via X11 forwarding (macOS requires XQuartz):
$ ssh -X [UnityID]@login.hpc.ncsu.eduVS Code:
While accessing Hazel through VS Code is possible, we recommend using either OnDemand or via SSH as these option are more user-friendly.
See the HPC VSCode guide
2.3 Your Spaces on Hazel
Once logged in, you have access to several distinct file locations. Each is sized and tuned for a different stage of your work — using the right one keeps your jobs fast and your group’s quota healthy.
You may notice below that some words below have a dollar sign in front of them. These represent varibles. When code is executed these placeholder variables will be replaced with the real data that they represent. For example $USER will be replaced with your unityid. You can confirm this by typing in the terminal echo $USER and pressing enter. Read more about variables in Command Line Basics.
| Location | Path | Quota | Best for |
|---|---|---|---|
| Home | /home/$USER |
1 GB/user | Shell config, small scripts, SSH keys |
| Scratch | /share/$GROUP/$USER |
20 TB/group | Active job working directory and pipeline intermediates |
| Applications | /usr/local/usrapps/$GROUP |
Varies | Conda environments, containers, shared software |
| RS1 | /rs1/researchers/... |
40 TB/project | Raw data, final results, anything you need to keep |
2.3.1 Home (/home/$USER)
Your personal landing spot when you log in. The 1 GB quota is small on purpose — home is for configuration, not data. Keep your .bashrc, SSH keys, and a handful of small scripts here. Don’t run jobs out of home or write large outputs to it; you will fill the quota and your shell will start failing in confusing ways.
2.3.3 Applications (/usr/local/usrapps/$GROUP)
Where your group installs shared software: conda environments, Apptainer containers, custom builds. Anyone in the group can load these in their job scripts, so you don’t each need your own copy.
Applications storage is read-only from compute nodes. Install or update software from a login node; jobs can only read from it.
2.3.4 RS1 (/rs1/researchers/...)
Long-term project storage. RS1 is where raw data lives and where final results go to rest. It’s slower than scratch but durable and not auto-deleted. Treat it as the canonical home for anything you’d be upset to lose.
Recommended workflow: Keep raw inputs in RS1 → stage them into scratch and run jobs there → copy final outputs back to RS1 → let scratch’s 30-day cleanup handle the intermediates.
2.4 Useful Commands
2.4.1 System and User Info
$ sinfo # available partitions and nodes
$ groups # groups you belong to
$ hostname # which node you are on2.4.2 Storage
$ du -sh . # size of current directory
$ du -sh /path/to/dir/* # size of each item in a directory
$ quota # home directory usage
$ quota -s -g [group_name] # group quota
$ ls -lrt # list files by modification time (newest last)2.4.3 Software
$ module avail # list available modules
$ module load [name] # load a module
$ which [executable] # path to an executable2.5 First Week Checklist
Before you start:
Getting connected:
2.6 Resources
- Getting started on Hazel
- Cluster status
- OIT HPC YouTube channel
- HPC announcements: join the HPC Announcements Google Group