Kubernetes supports two HTTP probes per container:
livenessProbe if it fails, K8s RESTARTS the container.
Use for "the process is wedged, kill it".
readinessProbe if it fails, K8s removes the pod from Service endpoints
but does NOT restart it. Use for "I'm warming up,
don't send me traffic yet".
Your app needs ~30s to load a model on startup. Best probe choice?
A) Aggressive livenessProbe with 5s timeout.
B) Readiness probe (so traffic isn't sent until model is loaded), and
a generous liveness probe with initialDelaySeconds.
C) No probes.
D) Both probes hitting the same /healthz with 1s timeout.Sign in to save your code and track progress across devices.
