How self-healing works
A ReplicaSet notices a missing Pod and replaces it within a second.
Scaling up and down
Change one number and watch Pods spread across Nodes.
Rolling update
Replace every Pod with a new version without downtime.
How a Service finds its Pods
Label selectors, EndpointSlices, and requests spread across ready Pods.
When a Node dies
Heartbeats stop, the Node goes NotReady, and Pods are evicted and rescheduled.
CrashLoopBackOff
A container that keeps failing is restarted with growing delays.
Readiness vs liveness
A failing readiness probe takes a Pod out of the Service; a failing liveness probe restarts it.
Autoscaling under load
The HorizontalPodAutoscaler adds Pods when CPU climbs, and removes them after five quiet minutes.
StatefulSets, in order
Pods with stable names, their own volumes, and a strict start and stop order.
Ingress routing
One entry point, many Services: ingress-nginx picks a backend by host and path.
NetworkPolicy
Default-deny cuts every connection to the Pods; an allow rule lets only the frontend back in.