Health endpoints

A service's health endpoints say whether it is alive and whether it is ready for traffic, and the platform's probes call them. Their paths and probes differ per service, so they are found in its code and deployment config rather than assumed.

Finding it

• The deployment config: livenessProbe, readinessProbe and startupProbe in Kubernetes manifests, a HEALTHCHECK in the Dockerfile, the load balancer's health check.
• The service's routes: /health, /healthz, /ready, or the framework's own (/actuator/health in Spring Boot).
• The port they answer on, which may differ from the service's main port.

Using it

• Call them directly after a start or a rollout: curl --fail --max-time 5 <url>/ready.
• Ask the platform as well: kubectl rollout status deployment/<name> waits until the new pods are ready.
• Readiness checks the service's own dependencies (its database, its queue); liveness does not, so an outage of a dependency does not restart every replica.

Examples: /health and /ready, Kubernetes liveness and readiness probes, Docker's HEALTHCHECK, Spring Boot Actuator (Java).