From the Notebook to the Cluster Workshop Series

workshop
reproducibility
Author
Affiliation

Erwin Lares

Research Cyberinfrastructure (RCI), Division of Information Technology, UW-Madison

Published

October 7, 2026

Modified

September 9, 2026

Abstract
If you have ever gotten an analysis working perfectly on your laptop and then watched it fall apart the moment someone else tried to run it – a missing package version, a path that only exists on your machine, an environment that quietly drifted out from under you – this series is for you. That gap between ‘it works for me’ and ‘it works, period’ is exactly where most computational research projects get stuck, and it is the gap this series is designed to close. From the Notebook to the Cluster is a three-part, hands-on workshop series for UW-Madison researchers who work in R and want to move an analysis from a personal laptop to the Center for High Throughput Computing (CHTC). It is not a series about three R packages. It is a series about a workflow: how to take a research project from ‘runs on my machine’ to reproducible, to portable, to something that can scale out across hundreds of CHTC jobs. I have written three R packages – toolero, containr, and submitr – that happen to make each stage of that workflow easier, and each session uses one of them as the vehicle. But the packages are in service of the workflow, not the other way around.

1 Roadmap

Each session picks up where the last one left off, following one research project through its evolution:

Local analysis
      │
      ▼
toolero    →   Build a reproducible project
      │
      ▼
containr   →   Encapsulate that project into a portable container
      │
      ▼
submitr    →   Run that container on CHTC

You will see this same project – a real, if modest, analysis – go from a folder of scripts on a laptop, to a well-organized and documented R project, to a container that runs identically on any Docker-compatible machine, to a job executing on CHTC and returning results. Each session solves a genuine problem that shows up somewhere in that arc, and each one hands you something you can keep using afterward.

2 The three sessions

All three sessions run from 2:00 to 3:30 PM in CS 3139 (Computer Sciences Building), in person. You are welcome to attend all three in sequence, or to attend only the one that addresses the problem you have right now – more on how that works below.

2.1 Part 1 – Starting Clean: Reproducible R Projects with toolero

Wednesday, October 7 | from 2:00 — 3:30pm | CS 3139

The starting point is the mess nearly every research project accumulates: scripts that only run in a particular order, dependencies nobody wrote down, a folder structure that made sense six months ago and no longer does. This session works through that mess directly, then introduces toolero to resolve it – reproducible project scaffolding, dependency management, and documentation practices that turn a personal working style into something someone else (including future you) can pick up and run. You will leave with minimal-project, a well-organized, reproducible R project that becomes the shared artifact for the rest of the series.

2.2 Part 2 – Portable Research: Containerizing R Projects with containr

Wednesday, October 14 | from 2:00 — 3:30pm | CS 3139

A reproducible project on your laptop is still tied to your laptop. This session asks why renv alone is not enough to guarantee that an analysis runs the same way somewhere else, and introduces containers as the answer. Using containr, you will build a Dockerfile, build and run an image locally, and publish that image to a container registry. You will leave with a published container image for minimal-project that runs identically on any Docker-compatible machine.

2.3 Part 3 – Scaling Out: Running Containerized R Projects on CHTC with submitr

Wednesday, October 21 | from 2:00 — 3:30pm | CS 3139

A portable container is ready to run anywhere – including, at last, on CHTC. This session covers what a containerized project needs to run as an HTCondor job: preparing and uploading the required assets, submitting the job with submitr, and retrieving results. You will leave with a working submission file, executable, and a completed CHTC job for minimal-project, along with the workflow to do the same for your own research.

3 Come to one, come to all

The series tells one continuous story, but you do not need to attend every session to get value out of one. Each session works toward a shared hands-on task, and if you do not finish yours in the time available, I will have a completed version of that session’s artifact ready so you leave with something real either way:

  • toolero: a finished minimal-project.qmd and its sample data, data/penguins.csv
  • containr: a finished Dockerfile, a built image, and that image published to a container registry
  • submitr: a finished .sub submission file, a .sh executable file, a submitted job, and its results

That means if you can only make it to the containr session, you will start from a working reproducible project (yours if you attended Part 1, mine if you did not) rather than getting stuck before the session’s real content even begins.

4 Who this is for

This series is aimed at researchers on campus who already write R code comfortably and intend to eventually run work on CHTC, but who have little or no experience with renv, containers, Docker, HTCondor, or CHTC workflows themselves. If that describes where you are, you are the intended audience – this is deliberately not a series for people who already have a container-to-cluster pipeline working.

5 What to bring

Getting these set up ahead of time means session time goes to the workflow itself rather than to installation troubleshooting.

6 When and where

Session Topic Date Time Location
Part 1 Starting Clean (toolero) Wednesday, October 7, 2026 2:00 - 3:30 PM CS 3139
Part 2 Portable Research (containr) Wednesday, October 14, 2026 2:00 - 3:30 PM CS 3139
Part 3 Scaling Out (submitr) Wednesday, October 21, 2026 2:00 - 3:30 PM CS 3139

7 Questions

Registration details are still being finalized. In the meantime, if you have questions or want to be notified when sign-up opens, reach out to me at erwinlares@wisc.edu .