– What exactly is Jupyter and Jupyter Hub?

– Students can access a fully working Python environment from their browser

– Setting up a fully working local development environment can be tricky

– 24 students (the size of the class) access the Jupyter Hub server

– Comparing Jupyter Hub to other alternative tools and services

– There’s a number of ways to do student authentication (oauth2) with Jupyter Hub

– Peter has been using this set up since the summer of 2018

– With Covid-19, now is a good time to be able to do remote work

– You don’t need to know much about how Jupyter Hub works to deploy it

– Sometimes your task is to deploy a code base you didn’t write

– Why you might want to self host things like Jupyter Hub and GitLab

– Initially Peter tried to code up his own solution but then found Jupyter Hub

– You need a Linux server, some type of auth, Python and a database to get going

– The server is hosted on a $40 / month DigitalOcean droplet

– You need a pretty beefy server even with relatively few users for Jupyter Hub

– There’s nginx acting as a reverse proxy in front of a Tornado web server

– Configuring nginx was Peter’s biggest pain point since he teaches Python not DevOps

– It was important to set up SSL certificates, he used Let’s Encrypt with Certbot

– From teaching Python to becoming Mr. DevOps (deployment is hard)

– The server is using Ubuntu 18.04 LTS and Peter connects using PuTTY on Windows

– Peter will try the next Ubuntu LTS in fall’s semester

– The server is running Python 3.7 to keep it compatible with Jupyter Hub

– Each quarter gets a brand new server that’s set up from scratch

– From manually configuring everything to Bash scripts to using Ansible for automation

– The Ansible set up has been used once so far

– Students never directly SSH into the server but it’s good to keep it locked down

– Peter can log into the server and check out his student’s work

– The next step for Peter’s automation project is to scale up the server’s size

– Other professors are using and planning to use this Jupyter Hub set up

– Peter usually updates the server’s packages when he turns it off to upgrade it

– Most instructors are using the college’s built in learning management system

– Jupyter Hub takes care of presenting logs and errors in its admin dashboard

– The college’s LMS is still the source of truth for looking at a student’s work

– Backups are done manually but Peter isn’t too worried about losing the data

– The Jupyter Hub is accessible on the public internet but no downtime alerts are used

– Best tips? Deployment is tough but it’s ok to make mistakes as you go

– Peter learned from many mistakes such as pushing SSH keys to GitHub and more

– After all is said and done, despite it being tough, it works well for his students

– You can find Peter on GitHub, Twitter and his blog, he also has an open source book