عرض عملي: إصدار Canary يلتقط انحدارًا

خطوة بخطوة

1. State the change on the order subsystem and the guard metrics with thresholds.
2. Deploy the new version at 5% traffic; show baseline and canary dashboards side by side.
3. Seed the regression (a slow query that raises p99 under load) and apply realistic traffic.
4. Watch the analysis fail at the first stage and traffic revert automatically; read the verdict.
5. Planned mistake: use a canary too small and too short to see the difference, so the bad version passes. Fix: raise volume and bake time to match the detection need.
6. Record what the run proved and what it could not.

اختبر فهمك