Teaching environment
For teaching we have a separate JupyterHub environment that contains the required packages needed for teaching basic Julia and Python Modelling. This document explains how to run the jupyterhub environment and how students can use it. Finally instructions for installing the environment are provided because it wasn't simple and I never want to have to figure it out again.
Using the JupyterHub for teaching.
This environment can be accessed here here:
Some (15) users have been set up by default and additional users can easily be added. If the class is larger than 15 students this will need to happen. Give each user a login and make sure every user has a different login otherwise they will interfere with each other. The usernames and passwords are given in the table below.
student1
teaching1
student2
teaching2
...
...
student15
teaching15
When the students log in they will see the context of three github repos have automatically been added to their folders:
https://github.com/SiFTW/MitchellLabTeachingMD is in a folder called MDModelling
https://github.com/SiFTW/MitchellLabTeachingCancer is in a folder called cancerModelling
https://github.com/SiFTW/MitchellLabTeachingPop is in a folder called popModelling
If you want to update you teaching material, just update the appropriate GitHub repo, as this will be substantially easier than updating each user's environment on JupyterHub.
Please note the project.toml/environment.yml are ignored and dependencies need to be manually added to the appropriate docker files downloaded when installing the environment (see below).
You now should be able to simply go to the correct folder for your teaching and each student should be able to run their sheet in their environment. Warning this environment is a transient Docker environment so any results they generate will be lost unless they download them.
Starting the JupyterHub for teaching.
The magic command is:
sudo docker run --name jupyterhubDockerTeaching -p9001:8111 --network jupyterhub_network -v /var/run/docker.sock:/var/run/docker.sock -v /data/jupyterhub:/srv/jupyterhub simon/jupyterhubmulti:latest
If you receive the error about the name being in use, just do:
sudo docker rm [ID of existing container from error message]
Installing the JupyterHub for teaching from scratch.
The process is based (loosely) on this page, with some modifications.
Pull the docker images we need:
docker pull jupyterhub/jupyterhub
docker pull jupyterhub/singleuser
Create a docker network that bridges the two images:
docker network create --driver bridge jupyterhub_network
Make the folders for jupyterhub
sudo mkdir -pv /data/jupyterhub
sudo chown -R root /data/jupyterhub
sudo chmod -R 777 /data/jupyterhub
Download the customised jupyterhub_config.py
wget https://mitchell.science/jupyterhub_config.py
Copy the jupyterhub_config.py to the correct location.
cp jupyterhub_config.py /data/jupyterhub/jupyterhub_config.py
Download the userlist and copy it into the correct location
wget https://mitchell.science/userlist
cp userlist /data/jupyterhub/userlist
Download the Dockerfiles for the singleuser (jupyter sheet) and multiuser (jupyterhub) environment.
wget https://mitchell.science/Dockerfile.multi
wget https://mitchell.science/Dockerfile.single
Build both the singleuser and multiuser docker images:
sudo docker build -f Dockerfile.multi -t simon/jupyterhubmulti .
sudo docker build -f Dockerfile.single -t simon/jupytersingleuser .
If you get permission errors on the singleuser Docker image you may need to edit Dockerfile.single to include the lines:
RUN mkdir -p /home/user/jovyan
RUN chown "jovyan" "/home/user/jovyan"
Right before the main julia dependencies are installed
Now run the magic command, and then login to the jupyterhub at http://139.184.170.218:9001/ :
sudo docker run --name jupyterhubDockerTeaching -p9001:8111 --network jupyterhub_network -v /var/run/docker.sock:/var/run/docker.sock -v /data/jupyterhub:/srv/jupyterhub simon/jupyterhubmulti:latest
Debugging the teaching JupyterHub
You might need to remove individual student's images as they won't reflect any changes otherwise:
sudo docker rm jupyter-student5
You might want to view the logs of the singleuser instance:
sudo docker logs jupyter-student2
All these docker images can quickly fill the SSD so consider:
sudo docker system prune
The config on JLP, and HPZ4 is different. You may have more success with these files for HPZ4:
wget https://mitchell.science/hpz4/jupyterhub_config.py
wget https://mitchell.science/hpz4/Dockerfile.multi
wget https://mitchell.science/hpz4/Dockerfile.single
wget https://mitchell.science/hpz4/userlist
The update script is also different (for some reasons the users have different names in this version of jupyterhub!):
wget https://mitchell.science/hpz4/updateGithubs.sh
this command works:
sudo docker run --rm --name jupyterhubDockerTeaching -p9001:8111 --network jupyterhub_network -v /var/run/docker.sock:/var/run/docker.sock -v /data/jupyterhub:/srv/jupyterhub simon/jupyterhubmulti:latest
If you get errors accessing user folders you need to create them:
mkdir /var/lib/docker/volumes/jupyterhub-user-student9/
then run the update script to fill it with content.
Updating the github repos in existing users
Once users have been created changes in the github won't be updated as each user's storage is cached. I have created a script that will go through each user's folder move the modelling folders to a folder called "oldFolders", and re-download the GitHub repos to get the latest changes. It can be downloaded and run as follows:
wget https://mitchell.science/updateGithubs.sh
chmod +x updateGithubs.sh
sudo ./updateGithubs.sh
You can update individual repos with individual scripts for each session.
Cancer:
wget https://mitchell.science/updateGithubsCancer.sh
chmod +x updateGithubsCancer.sh
sudo ./updateGithubsCancer.sh
MD:
wget https://mitchell.science/updateGithubsMD.sh
chmod +x updateGithubsMD.sh
sudo ./updateGithubsMD.sh
Population:
wget https://mitchell.science/updateGithubsPop.sh
chmod +x updateGithubsPop.sh
sudo ./updateGithubsPop.sh
Last updated
Was this helpful?