Testing in production: Ideas, experiences, limits, roadblocks
summary
Claims a runnable JS tool that simulates real users against cloud APIs, built with Node.js and Cucumber and request and deployed in Docker to exercise Dyson’s fleet of 1 million connected machines. It finds real problems: frequent runs cause DynamoDB throttling and API slowdowns, fake accounts fill databases and break pagination, so tests must be spaced, tagged via correlation headers to exclude metrics, and clean up generated data automatically.
AI-written summary · about
chapters (14)
- 0:00Intro
- 2:38Summary
- 3:32Why testing in production
- 8:23Why test in production
- 8:45Endtoend tests
- 12:15The solution
- 13:04Cucumber
- 13:45How it works
- 14:35Balance
- 18:22How
- 19:38Endtoend
- 19:59Fake data
- 21:48Clean data
- 22:41Recap
related talks
36mTesting In Production
43mBest practices for continuous testing and observability
32mDo Not Fight Your Testing Tools
25mTesting in Production
26mTales from the QA crypt
37mFront-end Testing for Skeptics
43mIntegration testing a cloud native application with JavaScript｜
11mCD Demands Continuous Testing: Why We Built a Testing Platform Layer on ECS Usin...