JupyterHub
Warning
Deploying JupyterHub requires an existing Kubernetes cluster.
Introduction
JupyterHub allows users to interact with a computing environment through a webpage. As most devices have access to a web browser, JupyterHub makes it is easy to provide and standardize the computing environment for a group of people (e.g., for a class of students or an analytics team). JupyterHub is a web-interface through which many "singleuser" Jupyter notebooks (one per user) may be launched.
Upon accessing JupyterHub users are presented with a number of "Profiles": pre-installed Python, Julia and R environments aimed towards different use cases.
The JupyterHub Kubernetes application is installed using the Platforms dashboard of Azimuth, onto an existing Kubernetes Cluster deployment.
Launch configuration
To get started, in the Platforms tab, press the 
New Platform button, and select JupyterHub.
You will then be presented with launch configuration options to fill in:
| Option | Explanation |
|---|---|
| Platform name | A name to identify the JupyterHub platform |
| Kubernetes cluster | The Kubernetes platform on which to deploy JupyterHub. If one hasn't already been created, check out the Kubernetes Overview. |
| App version | The version of the JupyterHub Azimuth application to use. |
| Notebook CPUs | The number of CPUs to allocate to each user notebook. |
| Notebook RAM | The amount of RAM to allocate to each user notebook. |
| Notebook storage | The amount of disk storage to allocate to each user notebook. |
Using JupyterHub
Accessing JupyterHub
After creating the JupyterHub platform, it will automatically be exposed by Azimuth's Zenith proxy.
JupyterHub can be accessed via the link under Services.
Users should be prompted to start a server, and be given three default options.
Note
Further development will allow for a custom image and/or a whole custom repo-to-docker codebase to be used
Profiles
| Profile | Description |
|---|---|
| Python environment (minimal) | A minimally functional Jupyter server; with basic commandline tools, a TeX environment and Python (along with the Conda and Mamba package managers). |
| Data Science Notebook | Contains popular scientific Python packages, the R interpreter and a basic set of data science packages, rpy2, and the Julia compiler along with notebook support and HDF5. |
| GPU-enabled Machine Learning environment | Contains GPU support and drivers; along with TensorFlow, Keras, PyTorch, and everything in the above Data Science Notebook. |
Warning
The GPU-enabled Machine Learning environment profile will only launch on clusters with a node group that has machines with a GPU-enabled flavour. It may take 10-15 minutes to launch this profile for the first time on a newly deployed GPU node. This is because Nvidia CUDA drivers are installed first automatically in the background, then the large ~8GB profile docker image needs to be downloaded and loaded.While waiting, you may see a message saying "No Capacity" or "No Nodes match the affinity/selector". This is okay, so long as there is a GPU node group available.
Server Hub
To switch profile or restart the code server from JupyterLab, head to File → Hub Control Panel.
From there, the server may be stopped and relaunched.
Managing Users
User management can be performed via the Identity Provider, Keycloak.
Access to JupyterHub can be granted by the platform or service specific Keycloak groups.