Deployment & Deliveryintermediate11 min · 9 steps
Health Checks
Tell the platform which instances are started, alive and ready — and tell on-call when the whole pipeline stops giving correct answers.
stale index found at09:00
signalrequesterrorresponse
metrics are illustrativeProblemStep 1 of 9
Every probe is green. The answers are wrong.
Overnight, the re-index job fails quietly and one region's knowledge-base articles vanish from the index. Every pod is running, every probe passes, the error rate is zero.
Users asking about VPN setup now get “I don't know”. The platform's health checks were never designed to notice — they ask is the process alive?, not are the answers right?
Watch outFor AI systems, “up” and “working” are different questions. You need checks for both.
Lesson 18 of 20 in Deployment & Delivery