I’m a Jupyterhub newb and am setting up a multi-team environment using “Zero to JupyterHub with Kubernetes” on Azure AKS. Do I need an instance per team, or will the Kubernetes deployment support multi-teams out of the box?
I’m also getting familiar with the JupyterHub and Kubernetes world, looking to do something similar (groups with shared filespaces, permissions etc., in my case classes with instructors and students).
The JupyterHub kubernetes deployment strategy doesn’t do teams out of the box. An instance per team is a strategy I’m looking at, but there are still some unknowns I’m starting to look into (Multiple hubs per kubernetes cluster?).
There have been quite a few creative deployments of JH on kubernetes described over at the gitter (https://gitter.im/jupyterhub/jupyterhub, where I just noticed you posted also ), but not many guides that I’ve found in the wild. Let me know if you figure anything out, and I’ll try to do the same!
BTW, for shared storage I’m looking at within-cluster NFS at the moment - maybe someday we’ll get to rook.io or similar, I’ve seen notes about that on gitter as well. This has been challenged by the k8s nfs volume provisioner not resolving DNS within cluster (see recent comments at https://github.com/kubernetes/kubernetes/issues/44528). But, the user images over at https://jupyter-docker-stacks.readthedocs.io/en/latest/ use a start.sh that runs when containers start; this runs scripts in /usr/local/bin/start-notebook.d - these are looking like a good place to do traditional NFS mount (which does work within-cluster, since the container itself is doing the DNS lookup). Finding a few bumps along the way, but nothing show stopping (yet).