Productionizing R scripts in the cloud
summary
Checklist for taking an R prototype to production: use git and code linters, pack reusable helpers into an internal R package, and enforce style via editorconfig and PR checks; add structured logging (example: the logger package with JSON layout and appenders) and central metrics; avoid plaintext secrets by using AWS KMS or environment variables managed outside code; run cloud ops via reticulate + boto3 for S3, schedule with Jenkins, and build reproducible Docker images pinned to an MRAN snapshot.
AI-written summary · about
chapters (17)
- 0:00Introduction to talk
- 0:45Why R in production
- 3:18Schedulers and tools
- 5:10Version control importance
- 6:34Coding with style
- 8:01Avoid repeating code
- 9:07Documenting and licensing
- 10:36Logging in R scripts
- 14:26Database connection security
- 18:27AWS KMS for secrets
- 21:03Database query functions
- 22:28RJava and JVM issues
- 23:11AWS SDK with reticulate
- 24:56Jenkins for scheduling
- 25:37Docker for reproducibility
- 26:43Pinning package versions
- 27:47Closing remarks
related talks
14mRobust R Deployments: Building a Pipeline from RStudio to Production
21mR in production
14mR Shiny - From Conception to the Cloud
19mConfiguration management tools for the R admin- 14mR Shiny - From Conception to the Cloud
01h 32mCreating Production-Ready Shiny Apps
17mR & Python playing nice, in Production
15mR Markdown + RStudio Connect + R Shiny