A quick-start list of commands for a fully-configured Jupyterlab user install that supports bash, R, and python.

See below for an updated version that uses docker!!

You can run these commands on a nice beefy server, or on your local machine. I prefer to use nice big servers, where data lives, not a laptop with limited compute & storage (unless your data needs are small). For WeigelWorld folks: that would be taco or chimi.

/usr/bin/python3 -m pip install -U pip
/usr/bin/python3 -m pip install -U jupyterlab nbconvert jupytext bash_kernel papermill
/usr/bin/python3 -m bash_kernel.install
Rscript -e 'install.packages("IRkernel"); IRkernel::installspec()'
jupyter server password # enter a NEW password here
tmux new-session jupyter lab --ip 0.0.0.0 --port YOUR_PORT --no-browser

Then you should be all set up. Go to your server’s address (like http://chimi.eb.local:YOUR_PORT or http://localhost:YOUR_PORT if running on your laptop) in a modern browser (Firefox/Chromium/Safari). It will probably ask for your password, which should be the one you set with jupyter server password, NOT your normal password for that server or laptop.

Some notes:

• You should pick a unique port number (between 1025 and 65000), which you will need to remember (or bookmark). Replace YOUR_PORT with your actual port number in the commands above and below.
• We run JupyterLab within tmux so that it will persist past the current ssh connection and we still can easily see its output. If you don’t know or want tmux, you could use screen or even nohup to get Jupyterlab to persist.
• For remote servers, a more secure alternative is to omit --ip 0.0.0.0 in the jupyter lab command above, and always use SSH forwarding. One would then use ssh -N -L YOUR_PORT:localhost:YOUR_PORT ldap_id@chimi.eb.local¹ (obviously replace that with your username and server), and access JupyterLab via http://localhost:YOUR_PORT, rather than the server URL above. JupyterLab and commands run from it will still run on the remote server, but you connection will now be encrypted and folks on the same network as the compute server won’t be able to access the JuptyerLab instance directly. This is very important if you run on an untrusted network or if running Juptyerlab e.g. in the cloud.

2024 update: Jetzt mit Docker!

After several annoying attempts at supporting recent R on old Ubuntu versions, I gave up and have containerised the below. On clusters/servers supporting either docker or singularity, the following should work:

apptainer exec \
	-B /tmp/global2 -B /ebio \
	docker://ghcr.io/kdm9/notebooks:latest \
	jupyter lab --no-browser --port $YOUR_PORT --ip 0.0.0.0

As above, set --ip and --port appropriately. The -B arguments map paths – please adjust these to include any directory you’ll need to access, including host /tmp filesystems, NFS mounts, and your home dir. Also include dirs like /opt if these contain software installs, e.g. an lmod module system tree.

This research was supported by a Marie Skłodowska-Curie Actions fellowship from the European Research Council.