1. First connect to O2 and open up an interactive session with all of the cores and memory you want to use. Here I’m connecting to the short queue so I can get more cores to use.

srun -n 8 --pty -p short --mem 64G -t 0-12:00 --x11 /bin/bash

1. Note the name of the compute node you are on:

uname --nodename

1. Start a jupyter notebook server on a specific port:

jupyter notebook --no-browser --port=1234

This command will open up a notebook server on port 1234. You might have to pick a different port if 1234 is being used. Note the token it provides for you, you will need this token to use your notebook server.

1. Create an auto-closing SSH tunnel from your local machine to the jupyter notebook:

On your local machine do:

ssh -f -L 9999:localhost:9999 o2 -t 'ssh -f -L 9999:localhost:1234 compute-a-16-49 "sleep 60"'

This sets up a two SSH tunnels. The first one is connecting port 9999 on your laptop to port 9998 on login02 on o2 (134.174.159.22). The second is connecting port 9998 on login02 to port 1234 on compute-e-16-49. This script will auto-close the tunnel if you don’t connect to it in 60 seconds, and will auto-close the tunnel when your session is closed.

1. Open a web browser and put localhost:9999 as the address.

This should now connect you to the jupyter notebook server. It will ask you for the token. If you put the token in, you can now log in and will be in your home directory on O2.

You are now running a notebook server. This is just running using a single core now– we want to hook up our computing that we reserved. We asked for 8 cores, so we’ll set up a cluster with 8 cores. Click on “IPython Clusters”, set the number of engines on “default” to 8, and you will have your notebook connected to the 8 cores.

1. Start working!

You can open up a terminal by going to the Files tab and clicking on new and opening the terminal. You can start a new notebook by going to the Files tab, clicking on new and opening a python notebook.