The Zombie Container

PagerDuty — 504s spiking, zero pod restarts

SRE: /api/checkout is timing out. Kubernetes shows 0 restarts — pods look healthy.

The trap: DB connection pool is exhausted. The Python process answers /health with 200 OK, so K8s keeps routing traffic to a deadlocked container.

Your task

Fix `health_check_endpoint()` in app.py — deep-check db.ping() and return 500 when the dependency is dead.

Book: add a health route · Prod: superficial liveness probes create zombie pods.