Improvements and Progress

Since winter quarter of 2020, the JupyterTeam has fixed a long standing issue where Thebe, the JavaScript package that we use to display and run code on libretexts.org, would not properly work with the entire interactive Jupyter widget ecosystem. By migrating from the Jupyter HTML Manager package to the Lab Manager, we are now able to run all ipywidget interactive widgets, which greatly expands the possibilities for textbook authors.

Example of an ipywidget controlling a matplotlib figure in Thebe.

In terms of cluster reliability, we have added a second router to be highly available with the already existing one; If the master router fails for any reason, then the backup one will take its place until the master comes back online. Furthermore, we configured the routers to act as TFTP servers so that they we may use Ubuntu netboot to load Ubuntu on all nodes from scratch. As compared to the previous method, this way is more effective and utilizes existing infastructure better.

We completed the migration of our default-env computing environment to JupyterLab 3.0, which includes new features which are great for widgets and their JavaScript extensions. Additionally, we added a new Adapt plugin to supplement our already existing Query plugin on libretexts.org.

In the Works

For the future, we are planning to develop a Ceph cluster to replace our NFS-client-provisioner as the storage solution for users on the JupyterHub. We also intend to extend the cluster with GPU nodes and provide more specialized computing enviornments. Feel free to use our Binder plugin on libretexts.org, and check out our JupyterHub if you are a UC Davis member!