Why it matters

Probes shape reliability. Understanding shapes ops.

Advertisement

The architecture

Liveness: still alive? restart if not.

Readiness: accept traffic? drain if not.

Startup: initializing.

K8s probe typesLivenessstill alive?Readinessaccept traffic?Startupinitializing?K8s probe design critical for self-healing + rolling updates; LLM startup slow
LLM health probes.
Advertisement

How it works end to end

Liveness restart on hang.

Readiness drains during load / init.

Startup for slow-init containers.