Liveness probe: is the process up and the async runtime responsive? This
deliberately does NOT probe the database. A liveness probe that fails on a DB outage makes the kubelet restart every replica, turning a recoverable dependency…
GET
deliberately does NOT probe the database. A liveness probe that fails on a
DB outage makes the kubelet restart every replica, turning a recoverable
dependency blip into a self-inflicted crash-loop. Readiness (
/health)
owns dependency health and load-balancer eviction; liveness only owns
‘should this container be killed and recreated’.