I Test In Production
summary
Argues that testing in production is inevitable and often superior, because staging never matches real users and emergent failures are common. Recommends shifting effort into high‑cardinality observability and using feature flags and canaries for progressive rollouts, with continuous load tests and beta cohorts (example: staged rollout that starts in a small region). Gives concrete tactics like replay 24h of production traffic for database upgrades, treats deploy scripts as production code, and cites an 80/20 bug split to justify reallocating engineering time.
AI-written summary · about
chapters (7)
- 0:00Start
- 6:09Definitions
- 10:15What Is the Complexity of Systems
- 19:48Chaos Engineering
- 25:56Test and Staging
- 26:27Real Risks
- 29:33Safety Nets
