Jupyterhub on DigitalOcean to serve notebooks for a course

Many of us use jupyter notebooks to teach classes and we often serve them through mybinder.org because all of our students experience the same environment no matter what platform they use. The only problem with binder is that any changes the students make to their notebooks are not recorded if their session times out - the student session on binder is not persistent across sessions.

Persistence does introduce some complications - every user has to be identified so that their session can be restored. Someone has to provide secure access and long-lived storage on top of the required computational environment.

This is an attempt to provide a template to build a cheap-and-cheerful jupyter-notebook server that can handle the needs of a medium sized class. Following this template means that the class coordinator can focus on the content rather than the setup.

Description

This repository can build itself into a digital ocean droplet that runs (the littlest) jupyterhub with the underlying dependencies that you specify in a conda environment yml file. The configuration of the droplet is stored in the repository secret variables. Github actions are configured to monitor the repository and update the content of the server when changes to the configuration are pushed. The actions also check to see if the server is alive and well.

This template repository contains sample content that runs on a dedicated droplet ... the status of that droplet is autoatically monitored:

How to use this template

The full set of instructions is given in the Documentation folder. The steps are:

• Create a suitable server (instructions are for a Digital Ocean droplet)
• Make a clone of this template repository (See Github.md)
• Update the cloned repository with information about the server that you have set up
• Commit the updates on Github to trigger the build / rebuild of the content of the server
• Create links for nbgitpuller for users of the server and test them (See nbgitpuller.md)
• Add your users to the jupyterhub (ManagingUsers.md)

Try out the Nbgitpuller

To make a "binder-like" link to a repository on a droplet that you have set up, you can read the nbgitpuller documentation or fill out a form here:

You can launch this example particular example to try it out by clicking on this link. Your work is persistent.

Administration tasks

If the hub has a signup page it can be reached here:

And the corresponding page for an admin user to authorise the users after they sign-up is

Admin users also have access to the hub control panel to shut down wayward servers and add / remove users.