Liveness, readiness, and startup probes

The kubelet can run three probes against a container. A startup probe verifies that the application has started; while one is configured, liveness and readiness probes do not run until it succeeds, and if it fails the kubelet kills the container, which is then handled by its restart policy. A liveness probe determines when to restart a container — for example, to recover from a deadlock — and must be configured carefully because an incorrect implementation can cause cascading failures; use initialDelaySeconds or a startup probe rather than waiting on readiness. A readiness probe determines when a container may receive traffic; on failure the EndpointSlice controller removes the Pod's IP address from the EndpointSlices of all matching Services, and it runs for the container's whole lifecycle. Scope: Kubernetes. Reviewed October 4, 2026.