Testing in Production
summary
Argues that staging lies and the only way to know a feature works is to test in production behind feature flags, using Split/LaunchDarkly/Harness; create test users (is_test_user=true), target automation bots with Robot Framework or Puppeteer, schedule runs via Jenkins/CircleCI, alert with PagerDuty/Slack, and monitor via Splunk. Deploy behind a production canary, run AAA tests first, require feature-flag cleanup, and note the real risk: misrollouts like Cloudflare’s 82% outage.
AI-written summary · about
chapters (16)
- 0:00Testing in production intro
- 2:04Staging environments fail
- 6:34Fear of production testing
- 7:56Feature flagging first tool
- 11:48Maintaining feature flags
- 13:13What to test in production
- 15:59Essential production tools
- 16:42Safe test data in prod
- 18:09Debugging production tests
- 19:10Test cleanup and alerts
- 20:14Canary deployments
- 21:36Monitoring production tests
- 22:19Mitigating production risks
- 23:43Shifting to prod testing
- 26:29Implementation guide
- 27:14Key takeaways & Q&A
