Initializing a Jupyter Notebook Instance

On ARC Sockeye, you have to initiate compute jobs using slurm, which is a command-line job manager.

Jobs are started with sbatch, monitored using squeue, and stopped using scancel.

You can view your jobs with squeue --user=$USER.

To start a job, run sbatch <script file>, which is typically a bash script.

For Jupyter Notebooks, I have copied a script from the ARC Confluence documentation. It lives at /arc/project/<allocation>/<CWL>/jupyter-datascience.sh.

What this script does is that it creates a new Jupyter Notebook instance on an ARC Sockeye compute node on an available port, then it produces a file with instructions on how to connect to that using your computer.

To run it and start a new Jupyter Notebook instance on ARC Sockeye, run:

sbatch /arc/project/<allocation>/<CWL>/jupyter-datascience.sh

What this will do is start a new job on ARC Sockeye, with the script that starts a new Jupyter Notebook instance.

Once this is run, it will produce a text file at:

connection.txt

This text document contains the previously mentioned instructions. Here's an example, your port will be different.

1. Create an SSH tunnel to Jupyter Notebooks from your local workstation using the following command:

ssh -N -L 8888:se238:52409 <CWL>@sockeye.arc.ubc.ca

2. Point your web browser to http://localhost:8888

3. Login to Jupyter Notebooks using the following token (password):

password

When done using Jupyter Notebooks, terminate the job by:

1. Quit or Logout of Jupyter Notebooks
2. Issue the following command on the login node (if you did Logout instead of Quit):

scancel 4889894

Here's a breakdown of the instructions:

• For point number one, this command should be run on your OWN laptop inside of a terminal window. DO NOT CLOSE THIS WINDOW AS IT ACTIVELY PROVIDES THE CONNECTION!

• Once this is run, go to your browser of choice, and navigate to http://localhost:8888