Stage 4: Flight Control — Know When to Wait, Help, and Land
A process that has started is not automatically ready for passengers. A Pod that is ready is not guaranteed to remain useful through every dependency failure. And a shutdown is not a moment; it is a bounded conversation between the workload, the kubelet, routing components, and in-flight requests.
Flight Control gives Apollo a vocabulary for those moments. We begin with the signals the kubelet observes, then move through the resource requests a scheduler uses, the pressure a node experiences, and the voluntary disruptions a budget can limit.
What you will understand
As you read, ask which component observes each signal, what action it may take, and how long that action is allowed to take. Also ask which failures a mechanism cannot prevent; a health check, resource setting, or disruption budget is never a general promise of availability.
When to take the controls
Use the Flight Control lab once you can say which actor reacts to each probe result and why a local successful rollout is evidence with limits. Then you can observe readiness, scheduling, and shutdown without overclaiming availability.