Jupterhub deployed via docker containers including pyiron package
This repository is forked from here.
The objective here is to create a dockerized jupyterhub which can spawn jupyter notebooks (with pyiron installed) as docker containers.
The structure of the repository is similar to the original repository. Therefore, for more details you can refer to the repository. Here, I try to explain what modifications have been made.
Creating the image of a single-user notebook
To build the single-user notebook images, first we use make notebook_image.
In the single-user/Dockerfile file, we included installation of pyiron via conda-forge channel.
Building the docker-compose.yml
In the second step, the hub image and the database image is created. By performing make build, beside creating docker volumes and copying several files (secrets, userlist, jupyterhub_config.py) to the created image, the docker-compose.yml file is used to create the image for the hub and the database.
The jupyterhub_config.py file sets the authentication class to PAM authentication from the host machine. Additionally one needs to mount /etc/pass and /etc/shadows on to the image. This is implemented in the docker-compose.yml file.
There are minor changes to the Dockerfile.jupyterhub file to include conda package in the docker image.