Pritykin Lab Onboarding

Ask Yuri for:

1. keycard access to comp space
2. Slack access
3. Argo access
4. Github access

Set Up Argo

1. Download iterm (some nice features but not necessary)

Using Argo

Login. Open iterm. Run ssh {computing_id}@argo.princeton.edu

1. If you see something about adding a key, just say yes (this just happens on the first try
2. Password instructions through email

Using the Terminal

1. Know some unix commands. I would not memorize commands, just do what you can to get the job done. Know stuff about directory navigation everything else is extra: https://www.geeksforgeeks.org/linux-tutorial/

VSCode

If you don't know how to create a file in terminal, install VSCode.

1. Opne VSCode
2. Go to the extension icon on the left then search for Remote - SSH
3. Download Remote - SSH
4. Go to the bottom icon with a computer and some circle on the bottom right
5. Add ssh host (there is plus button next to window)
6. New popup and type in you username and password as done for terminal
7. choose the top option for adding a configuration

Set up Miniconda on Argo

Read a little about why you would want to create environments: https://conda.io/projects/conda/en/latest/user-guide/getting-started.html

Miniconda is the primary way to download packages into a self-contained environments.

https://docs.anaconda.com/free/miniconda/#quick-command-line-install

Go to command line setup and run these sort of commands (find the newer ones):

mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm -rf ~/miniconda3/miniconda.sh

Reset terminal:

conda create -y -n jupyter
conda activate jupyter
conda install anaconda::jupyter

More generally, if you want to create a new environemnt (and this is the first thing I do on a new project/type of script):

conda create -n {name_env}
conda activate {name_env}
conda install pandas # search on google the package you want to install then install it

Inside base you can run:

conda install conda-forge::mamba

Mamba is a faster version of conda. You would then just use commands like mamba install x

Running Jupyter Notebooks

Most work can be done by starting a jupyter notebook on a slurm node using run_jupyter.slurm. Make sure to run have a conda environment with jupyter running.

conda activate jupyter
sbatch run_jupyter.slurm

After the sbatch you will get a file run_jupyter.out. Open the file to get a command like this:

Command to create ssh tunnel:
ssh -N -f -L 8890:argo-33:8890 [email protected]

Use a Browser on your local machine to go to:
localhost:8890  (prefix w/ https:// if using password)

On your terminal without connecting to argo access:

ssh -N -f -L 8890:argo-33:8890 [email protected]

In your browser then got to https://localhost:8890

Notes about Slurm Jobs

See your slurm jobs:

squeue -u dl4257

Example headers:

#!/bin/bash
#SBATCH --nodes=1                # node count
#SBATCH --ntasks=1               # total number of tasks across all nodes
#SBATCH --cpus-per-task=2        # cpu-cores per task (>1 if multi-threaded tasks)
#SBATCH --mem-per-cpu=8GB         # memory per cpu-core (4G is default)
#SBATCH --time=24:00:00          # total run time limit (HH:MM:SS)
#SBATCH --mail-user={}@princeton.edu
#SBATCH -o main.out

FAQ

You can't connect to a jupyter server because address is already in use:

bind [127.0.0.1]:8890: Address already in use
channel_setup_fwd_listener_tcpip: cannot listen to port: 8890
Could not request local forwarding.

Answer: Rerun ssh -N -f -L 8890:argo-11:8890 {id}@argo.princeton.edu after failing to open the url.