Notes for setting up JupyterHub on localhost/server with custom authentication, and using Dockerspawner to provide each user with a separate container.
- Setup JupyterHub
- Add Custom Authentication
- Provide a pre-built test notebook (read-only) to each user
- Provide a separate Docker container to each user
- Integrate Docker Swarm
To run JupyterHub locally, you can use Vagrant. Simply running vagrant up --provision shall setup the base box, and provision it. After which you can move to installing Docker.
To setup on a server, run provision.sh. After that install docker.
Docker Documentation for Installation
After that, you can either:
- Pull the docker image for jupyter via:
docker pull jupyterhub/singleuser:latest - Clone the Dockerspawner repo & modify their Dockerfile & support scripts
We've taken Dockerspawner's Dockerfile and updated it to include the a sample notebook inside the container. This notebook shall be a read-only notebook, and shall be accessible to every user in their respective container. Similarly, files can be added
Navigate to the project directory as root user, and start Jupyterhub by running:
jupyterhub -f PROJECT_PATH/jupyterhub_config.py
Sincere thanks to all the contributors of Project Jupyter & JupyterHub, and all the people who documented their use cases & deployment notes for JupyerHub
- JupyterHub Docs
- JupyterHub - Github
- JupyterHub - Dockerspawner
- JupyterHub Tutorial: I highly recommend going through this, along with the accompanying video
- JupyterHub: Deploying Jupyter Notebooks for students and researchers
[VIDEO] - JupyterHub-Swarm: Running JupyterHub with Docker Swarm