Jupyterhub in the Classroom

​

Sang-Yun Oh

Data Science Instructional Computing Working Group

Who to contact

Ernesto Espinosa: cluster administration

• Create/manage Jupyterhub cluster
• Monitors credit usage and archive persistent storage for 1 year

Patrick Windmiller: instructor liaison

• Liaise between LSIT and the instructor
• Customize computing environment for each course

Sang-Yun Oh: faculty advocate

• Teaching with Jupyterhub (tools for submission, grading)

What is Jupyterhub?

• Jupyterhub is a multi-user Jupyter notebook server
• Jupyterhub authenticates users and instantiates notebook sessions
• All users get the same environment: “single-user notebook” (SUN)
• Jupyterhub needs Docker image SUN (modify Dockerfile for customization)
• SUN can be any image here: https://github.com/jupyter/docker-stacks
• Custom R images are created from: https://github.com/rocker-org/rocker
• Jupyterhub cluster runs on GCP and SUN images are on Docker Hub
• Persistent storage

Requesting Jupyterhub Setup

1-2-3!

1. Request Google Cloud Credit (more on this later)
2. Contact Patrick to discuss your needs:�E.g. Python vs. R studio, package needs, etc.
3. Contact Patrick and Ernesto for support�Contact Sang about questions regarding grading tools, etc.

Requesting Cloud Credit

• https://edu.google.com/programs/credits/teaching (example on next slide)
• Redeemable right away, expires 1 year after course start date
• Google provides coupon codes totaling (up to $5000)� $10 x number of students + $50 x number of teaching staff
• Additional requests are allowed
• For Jupyterhub, lump sum credit is needed

Sample GCP Credit Request Form Data

​

• First Name: Sang-Yun
• Last Name: Oh
• Email Address: syoh@ucsb.edu
• Title: Prof.
• Name of School: UNIVERSITY OF CALIFORNIA SANTA BARBARA
• Country: US
• Name of Department: Statistics and Applied Probability
• Link to Faculty Directory: https://www.pstat.ucsb.edu/people
• Expected Start Date: 2019-09-25
• Name of course/project: PSTAT 134
• Course abstract or project description: Overview and use of data science tools in R and Python for data retrieval, analysis, visualization, reproducible research and automated report generation. Case studies will illustrate practical use of these tools.
• Student email domain: ucsb.edu
• Expected Number of Students: 85
• Expected Number of TAs and Staff: 7
• Additional Comments: We create a Jupyterhub cluster for this class. We need *one lump sum credit* to setup and run this Jupyterhub cluster. During a lecture, students will be informed what Google Cloud Platform is and how it powers their JupyterHub.

Informing Students about GCP

Background: Google Cloud is increasingly wanting students to get direct exposure to GCP. Feel free to use the following slides in your lecture

• Jupyter Notebook on GCP
• RStudio on GCP
• Google Cloud Platform Solutions

JupyterHub URL: https://[CLASS].lsit.ucsb.edu

Internet

Amy

Sam

Amy’s Kernel

Sam’s Kernel

JupyterHub URL

JupyterHub URL: https://[CLASS].lsit.ucsb.edu

Internet

Amy

Sam

Amy’s Kernel

Sam’s Kernel

JupyterHub URL: https://[CLASS].lsit.ucsb.edu

​

[Image Source: https://rb.gy/cknf59]

​

Sample GCP Credit Request Response

Clipped email response from Google containing the credit code (forward to Ernesto):

Tools for Teaching with Jupyterhub

Jupyterhub Admin Interface

Moving Material to/from Jupyter Notebook

Graphical User Interface (GUI)

• Upload/Download buttons
• Drag-and-drop upload files (Jupyter lab)
• Clone a Github repo [Test Link]:�https://jupyterhub.github.io/nbgitpuller/link

Command Line Interface (CLI)

• `wget` or `curl` from direct downloadable links: e.g. Box direct link
• `zip directory.zip -r directory/` then Download

Moving Material to/from R Studio

Graphical User Interface (GUI)

• Upload/Export buttons
• Uploading a zip file expands into directory
• Exporting a directory creates a zip file

Command Line Interface (CLI)

• `wget` or `curl` from direct downloadable links: e.g. Box direct link
• `zip directory.zip -r directory/` then Download

​

Questions from Students

• Q: How do I delete a directory?�A: After emptying the directory
• Q: How do I download a directory?�A: Either zip (CLI) and Download or select all files
• Q: I can’t save my file! What might be the cause?�A: Storage allocation may be full or Jupyter notebook file may be too large

Inconveniences

• Access control is done manually�Default is to allow any UCSB ID
• There is no collaboration functionality�But instructor can assume any user identity via Admin panel
• No easy way to submit homeworks to GS for students�Manually download from Jupyterhub then Upload to GS
• No easy way to monitor load (need GCP login)�Potential for abuse of (limited) resources

Future Plans

• Possibly using AWS (year long credit)
• Tutorial on creating homework assignments: jassign
• Integration with GauchoSpace? e.g. access control
• Wifi reliability testing in large classrooms�(a tip: can use Zoom screen sharing if classroom projector is dim)
• Infrastructure developments: e.g. help-ticket system