Skip to content

Looking Around

Goal: Give you a fast, safe way to poke around the Betty cluster and discover what’s available—without going deep. Each section ends with a pointer to the right, longer guide.

Prerequisites: Make sure you have an account and have gone through the Logging-In tutorial

Who this is for?

  • You have a Betty account and can SSH into the login node.
  • You want a light-touch orientation: where files live, what software exists, how to see queues/GPUs, and how to run a tiny test job.

Rule #1: Do not run heavy compute on login nodes. Use Slurm (srun, sbatch, salloc) for anything beyond quick inspection.

1) Log in & orient yourself

From your laptop/terminal:

kinit <YourPennKey>@UPENN.EDU
ssh <YourPennKey>@login.betty.parcc.upenn.edu

Read the login banner/MOTD for maintenance notices or policy reminders. You will see the storage that you have access to as part of the login banner. This will show you where the storage is (path) as well as how much has been used. If you see the State as anything other than “OK”, you should work with your lab to investigate what is taking up so much space or so many files.

-----------------------------------------------------------------------------------------------
                                        Storage Quotas
-----------------------------------------------------------------------------------------------
| Path                            | Used      | Limit     | INodes Used | INode Limit | State |
| /vast/home/c/chaneyk            | 41.02 GB  | 50.00 GB  | 122.04 K    | 250.00 K    | OK    |
| /vast/projects/chaneyk/test     | 104.86 MB | 5.00 TB   | 3.00        | 25.00 M     | OK    |
| /vast/projects/jcombar1/testing | 27.85 TB  | 500.00 TB | 83.61 M     | 2.50 G      | OK    |
-----------------------------------------------------------------------------------------------
-----------------------------------------------------------------------------------------------

After logging in, you can start to take a look at where you landed:

whoami
hostname
pwd
date
  • whoami Your username is your identifier cluster wide, fortunately since this is your PennKey, it should be easy to remember.
  • hostname The login node where you landed. We run multiple login nodes, if you ever use a tool like tmux, you will need to come back to the same node so this is useful to keep in mind.
  • pwd Your present working directory. When you first login, this should always be your home folder. This is where all of your code and configurations will live. Small datasets can also live here.
  • date This will show you the current time along with the timezone. If you are working somewhere else, always keep in mind what timezone the cluster uses when you are tracking down jobs or new files!

A sample set of the information can be seen below:

chaneyk
login01
/vast/home/c/chaneyk
Fri Sep 12 02:49:05 PM EDT 2025

Learn more: See the Zero‑to‑MPI on Betty guide (SSH + first steps).

2) Where are my files? (Home and Projects)

Typical layout on Betty:

  • Home: backed by high‑performance storage (for configs, light data).
  • Projects: shared space for your research group.
    • VAST storage

Quick checks of your home folder, these quick commands will navigate you to your home folder and then show you some information about it:

cd ~
pwd
ls -lah ~

We have some helper functions to provide some quick context on what storage you have and finding what is taking up so much space:

parcc_quota.py

You will see paths print out similar to what you see when you first login. You can list out storage amounts within this folder. This is particularly useful when your quota is full:

parcc_du.py /vast/projects/<your-project>

Data Transfers

To quickly transfer data, you can use common commands to send data directly with the login nodes:

scp <your-local-file> <PennKey>@login.betty.parcc.upenn.edu:/vast/projects/<your-project>

This will use ssh authentication to help you with transferring your file.

Learn more: See Storage & Quotas on Betty (paths, best practices, cleanup).

3) What software is installed? (Modules & Python envs)

Betty uses environment modules (Lmod) to provide quick and easy access to commonly used software packages. To take a quick look through the very long list of pre-installed software or look for something specific:

module avail                 # browse what’s on the system

module spider anaconda3      # deep search (e.g., module spider cuda)

We will be looking at anaconda3 so that we can also look at python environments. Once you have verified that the software you are looking for is on the system, you can show what it sets and then load it:

module show anaconda3        # what exactly a module sets
module load anaconda3

Now that you have anaconda3 loaded, you have access to the conda command. This will let you manage your environments and any specific packages you may need.

# Conda style
conda env list

conda create -n tutorial

Best practice: keep project‑specific python packages in your own envs; don’t ❌❌ pip install --user into system Python ❌❌

Learn more: See Python & Conda/Mamba on Betty and Software Modules (Lmod) Basics.

4) What queues exist? Are GPUs available? (Slurm peek)

Cluster overview command such as sinfo and parcc_free.py allow you to get a sense for what resources are available at any moment in time.

sinfo  # partition, avail, time limit, node cnt, state, node list
parcc_free.py # Partition, nodes, CPUs, Memory, GPUs, Down

Your jobs & the system’s pulse:

squeue -u $USER   # What jobs you are running
squeue | wc -l    # How many total jobs are running

One safe “hello GPU” test (runs on a GPU node, then exits):

srun -p dgx-b200 --gpus=1 -t 00:01:00 nvidia-smi

Learn more: See Slurm Basics on Betty (partitions, accounts, QOS, job examples).

5) Good citizenship (quick reminders)

  • Login nodes: edit, sync, submit, not train.
  • Use project storage for shared data.
  • Small interactive allocations are great for debugging; release them when done.
  • Prefer containers/conda envs over system‑wide installs.

Learn more: See Betty Usage Policies and Data Management on Betty.

6) Troubleshooting starters

  • Command not found? → module spider <tool> or load the right module.
  • Job won’t start? → wrong account/partition/QOS, or resources not available. Check squeue, then scontrol show job <JOBID>.
  • Disk full? → run parcc_du on the path; purge caches/temp files; move to project storage.

Learn more: See Slurm Troubleshooting and Storage & Quotas.

7) Where to go next

Ask your PI or project admin for your correct account, project path, and preferred partitions. If anything looks off, open a support request via the PARCC site.