Saturday 16:30–17:05 in Megatorium

Deploying JupyterHub with Kubernetes

Camilla Montonen

Audience level:


JupyterHub is a great way to provide a data analytics environment for a class, a research group or a team of data scientists. In this talk, we will go through JupyterHub: what it is and how it works and how to setup a production ready deployment on a cloud provider using Kubernetes as the underlying infrastructure.


The main aim of this talk is to introduce developers, data scientists and data team leads to the capabilities of JupyterHub (what problems can it solve) and how to best deploy and manage it in production on a cloud provider. No previous knowledge or JupyterHub internals or experience in running it on cloud providers is required so this talk is suitable for complete novices.

The talk will be divided into four parts:

Introduction to JupyterHub

In this section, I will give a brief overview of how Jupyterhub and how it can be used to provision Jupyter notebooks for classroom or industry use.

A brief tour of JupyterHub internals

In this section, we will explore the components that make up a JupyterHub deployment: the HTTP proxy, the Hub and the Jupyter notebook servers for individual users We will use the insight we gain in this section to understand how to run a production ready JupyterHub deployment on a cloud provider's infrastructure.

Running JupyterHub in the cloud with Kubernetes and Helm

Now that we understand the basics of how JupyterHub works, we can deploy it to a cloud provider (Google Cloud will be used as an example throughout the talk) with Kubernetes. Never heard or Kubernetes or not sure what value it can add to a JupyterHub deployment? Not to worry - before we a do a deep dive into the deployment demo, we will briefly go over what problems Kubernetes helps to solve and how it does this. To wrap up this section, we will talk about making JupyterHub deployments easier with the help of Helm charts and what are some known pitfalls in running JupyterHub in production.

Security and Cost

Last (but definitely by no means least ) is making sure that our JupyterHub deployment is secure. This means ensuring that clients communicate with the Hub using HTTPS and that authentication is properly secured. Finally, I will compare the costs of hosting JupyterHub deployments on AWS, Google Cloud and Azure.

Subscribe to Receive PyData Updates