Testing in Production
summary
Argues that staging is unreliable and recommends testing in production by deploying code behind feature flags, targeting internal users (example: automationbot1) or mocking flags with three explicit tests (flag on, flag off, neutral), marking test objects with is_test=true so analytics separate test from real traffic, and using percentage rollouts or a canary release plus A/A tests. Recommends tools: Split (split.io) for flags, Robot/Cypress/Puppeteer for automation, Jenkins/CircleCI for scheduling, and PagerDuty/Slack for alerts. Start small and automate.
AI-written summary · about
chapters (11)
- 0:00Introduction to testing in production
- 1:28Problems with staging environments
- 7:02What is testing in production
- 8:05How feature flags work
- 10:15Automating tests with feature flags
- 13:46Managing test data in production
- 16:11What to test in production
- 19:25Mitigating risks of testing in production
- 21:08Benefits of testing in production
- 22:58Overcoming fear of testing in production
- 24:23Q&A session
