zach-schoenberger / CS6460

CS6460

This is the python and helm chart definition for my project JupyterAssessor for CS6460. This uses interacts with the code in jupyter-spawner to dynamically spin up testing pods for JupyterHub.

Setup

First setup kubectl to connect to your kubernetes cluster. This depends on your cloud provider or kubernetes setup. An example can be found here. Then checkout the repository and change directories.

git clone https://github.com/zach-schoenberger/CS6460
cd ./CS6460

Then run the helm-install.sh script to install the helm charts.

./helm-install.sh

You should now be able to run kubectl --namespace jhub get svc and get an output that looks like:

NAME              TYPE           CLUSTER-IP      EXTERNAL-IP      PORT(S)                      AGE
hub               ClusterIP      10.47.244.40    <none>           8081/TCP                     21d
jupyter-spawner   ClusterIP      10.47.240.248   <none>           80/TCP                       21d
proxy-api         ClusterIP      10.47.253.14    <none>           8001/TCP                     21d
proxy-public      LoadBalancer   10.47.253.103   34.66.49.226     80:30560/TCP,443:31998/TCP   21d
redis-master      LoadBalancer   10.47.240.132   35.224.232.159   6379:32590/TCP               40d
redis-slave       ClusterIP      10.47.253.0     <none>           6379/TCP                     40d

You should see at least all of the services above. Once the environment has finished spinning up, the EXTERNAL-IP of proxy-public is the ip to be for the JupyterHub login.

Customization

Notebook Container

To customize the container that runs the notebooks, modifications should be made to the jhub-spawner-client/jhub.Dockerfile. This is where packages should be added or removed, along with any other changes. If there is a default set of files that should be made available this is where to do it. The change the build.sh file to tag the new image with your image tag. Then modify the jhub-config.yml file to use the new image that was created. The singleuser and spawner image configs should be updated to use the new image.

Assessor Customization

This is the aspect that teachers always need to customize. To customize the assessor, the teacher should modify the python script that is used to run the students submissions to assess and grade them. Inside jupyter-spawner the file createAssessor.sh is used to load the assessor python file into Kubernetes as a configmap. There is an example default script in the repository called defaultPyscriptAssessor.py that simply runs the submitted script.