عرض عملي: تجربة فوضى بفرضية ومعايير إيقاف

خطوة بخطوة

1. Pick the question: does checkout still succeed if the payment provider is slow?
2. Write the hypothesis, steady-state metric (checkout success rate), and fault (added latency).
3. Set blast radius (one percent of traffic or a test cohort), abort criteria, and who watches.
4. Run, observe, and abort if the criterion is hit.
5. Planned mistake: start in production at full traffic with no abort condition. Fix: start in a pre-production environment, then a small cohort, with explicit abort conditions.
6. Hold the review: findings (missing timeout), actions, owner, and a repeat date.