JupyterHub from the Ground Up with Kubernetes
summary
Explains how to deploy a secure, production-ready JupyterHub on Kubernetes using the official Helm chart, shows the hub + configurable HTTP proxy + per-user notebook servers model, and demonstrates a running deployment (recorded demo) on a 3-node cluster. Covers authentication, per-container CPU/memory limits, persistent volumes, autoscaling nodes, and cleanup of idle servers. Notes practical cloud experience: Google Kubernetes Engine easiest, small demo cost ≈ £70–82/month, AWS requires more bootstrapping.
AI-written summary · about
chapters (18)
- 0:00JupyterHub introduction
- 1:13Live demo on Kubernetes
- 3:02Jupyter history and ecosystem
- 5:50Scaling Jupyter challenges
- 7:16How JupyterHub works under the hood
- 10:30Live demo internals on Kubernetes
- 13:31Production challenges
- 15:19Containerization role
- 17:45Kubernetes orchestration
- 19:31Why Kubernetes for JupyterHub
- 21:40Deploying with Helm
- 23:03Testing on cloud providers
- 27:16Choosing a cloud provider
- 28:40Summary
- 29:25Q&A 1
- 31:38Q&A 2
- 33:24Q&A 3
- 35:12Q&A 4
related talks
27mDeploying JupyterHub with Kubernetes
19mJupyterHub: A "things explainer overview"
39mHow We Hosted 200+ Data Scientists in a K8S Cluster
35mLarge Scale Teaching Infrastructure with Kubernetes
24mJupyter Notebooks for Ops
19mOptimizing Resilience and Availability by Migrating from JupyterHub to the...
30mOnyxia: A User-Centric Interface for Data Scientists in the Cloud Age
35mOptimizing Kubernetes Deployments with Helm