Kubernetes uses two types of health checks with different purposes.
Liveness probe:
"Is the process alive?" Failure triggers container restart.
- Check: Simple endpoint, no dependencies
- Example:
/healthz/livereturns
Readiness probe:
"Can this instance serve traffic?" Failure removes from load balancer.
- Check: Verify dependencies (database, cache)
- Example:
/healthz/readychecks connections
Key rule: Never check external services in liveness probe. Cascading restarts will destroy your cluster.