Ansible role that installs a public-facing Jupyter instance with JupyterHub and JupyterLab on Ubuntu
Go to file
Taha Ahmed 3d32fc6c75 Install Julia kernel for every user. 2023-08-21 07:39:14 +02:00
defaults Better notebook_dir handling 2023-08-21 04:04:29 +02:00
files Role variables now more self-contained (less dependency on group_vars) 2022-07-12 06:09:22 +02:00
handlers Discovered some syntax errors. Fixed. 2022-07-27 20:59:54 +02:00
meta Allow dict variable in defaults to merge (be overridden) by 2023-08-20 23:59:20 +02:00
tasks Install Julia kernel for every user. 2023-08-21 07:39:14 +02:00
templates Install Julia kernel for every user. 2023-08-21 07:39:14 +02:00
.gitignore Moved all config files to a single location. Stepping stone 2023-04-17 20:24:23 +02:00
LICENSE Initial commit 2021-10-02 00:01:13 +02:00 Full install inside python venv works. Resolves issue #2. 2023-04-18 02:54:16 +02:00

JupyterLab and JupyterHub

This Ansible role installs JupyterLab behind JupyterHub all inside a Python virtualenv. Tested on Ubuntu server 22.04.

This is the actual code I use to maintain my own Jupyter server. The code should be workable as-is for others, as long as you reset the role variables and create your own cookie secret.

If you face problems, contact me directly or open an issue, and I'd be happy to help you!

Dependencies (other Ansible roles)

NOTE! The R role has dependencies of its own. If you want the R kernel, the R role presently depends on TeXLive, which is a rather large download and installation. I suggest you keep the R kernel disabled, unless you are willing to either install TeXLive or edit the R role manually to remove that dependency.

Example playbook

This section is written with an Ansible beginner in mind.

This is a very rudimentary example. The web contains plenty of guides on how to write Ansible playbooks.

In short, you will need to install Ansible, then create a directory to hold your playbook and create the following files in it: hosts.yml, ansible.cfg, and playbook.yml. Here are some working examples to get you started (replace all chevron-enclosed strings, <>, with your own):

### hosts.yml
    ansible_python_interpreter: /usr/bin/python3
      ansible_host: <hostname>

ansible_host is where you tell Ansible how to reach your host. It can be an IP address, but I suggest you learn to use SSH and ~/.ssh/config. Then you can just use the same Host-name in your Ansible hosts file, and Ansible will obediently follow your ssh configuration.

It's also possible to install Jupyter etc. on the same machine where you have installed Ansible (you will want to search for "running Ansible playbook on localhost").

### ansible.cfg
inventory = ./hosts.yml
become_method = sudo
### playbook.yml
- hosts: <hosts-groupname>
  become: yes
    - { role: jupyter }

Make sure you download this Jupyter role and all its dependency roles into the folder roles/ inside your playbook directory. I suggest you simply git clone each role from inside your roles/ directory. You should end up with a tree looking something like this:

├── ansible.cfg
├── hosts.yml
├── playbook.yml
└── roles
    ├── jupyter
    ├── python3
    └── nodejs

Finally, before running the playbook, please generate a cookie secret for Jupyter by entering the ./roles/jupyter/ directory and saving a random string to the files subdirectory, like this:

cd ./roles/jupyter
sudo openssl rand -hex 32 > files/jupyterhub_cookie_secret

Now run your playbook (make sure you are in your playbook's top directory):

ansible-playbook playbook.yml --ask-become-pass

Since this role installs system packages, you will need to pass your sudo password for the machine Ansible is targeting to the playbook. This is what --ask-become-pass does--it makes Ansible ask for your password. If you don't provide a password (or the wrong password), the playbook will fail at the first task that requires elevated rights (but not before).

Known issue

Upon successful login, we get stuck on a loading page saying "Your server is starting up. You will be redirected automatically when it's ready for you", but it never does. Refreshing the page (F5) or clicking the "Home" link or navigating using the visible JupyterLab UI takes us past this page, and then everything works normally. Not sure why the redirect to lab/ does not work (haven't really investigated this issue).

Jupyter and its subprojects

Julia kernel

Other kernels