JupyterHub Docker

Overview

JupyterHub Docker is a Docker image for JupyterHub. A default authentication for JupyterHub Docker uses GitHub's OAuth.

Build Docker Images

Ubuntu 16.04 based Image

./bin/jupyter-docker build -t ubuntu18.04

Machine Learning Image

./bin/jupyter-docker build -t machinelearning

CUDA based Image

./bin/jupyter-docker build -t cuda10.0-cudnn7

CUDA based Machine Learning Image

./bin/jupyter-docker build -t cuda-machinelearning

Setup

OAuth (GitHub)

Create a configuration file:

cp env/localhost env/`uname -n`
cat env/`uname -n`
CONTEXT_PATH=/
GITHUB_CLIENT_ID=<<GitHub Client ID>>
GITHUB_CLIENT_SECRET=<<GitHub Client Secret>>
OAUTH_CALLBACK_URL=http://localhost:8000/hub/oauth_callback

For GITHUB_CLIENT_ID and GITHUB_CLIENT_SECRET, see Building OAuth Apps.

Login Users

JupyterHub Docker uses acceptable users list, res/userlist. You need to update userlist file. The format is <userid> <username> <mode>. userid is an user id in Docker container, username is GitHub account, and mode is admin or (empty). For details, see res/addusers.sh.

Operations

Run Docker Container

./bin/jupyter-docker run -t <tag name>

Tag names are as below:

ubuntu16.04
cuda10.0-cudnn7
cuda-machinelearning

URL for this Jupyterhub is http://localhost:8000/.

Stop Docker Container

./bin/jupyter-docker stop -t <tag name>

tohtsky/jupyterhub-docker