Foundations

Why foundations matter

Kubernetes punishes vague mental models. Most operational damage comes from acting on assumptions that do not match the control plane’s real semantics.

This shelf is written to stabilize your internal map: what is authoritative, what is eventual, what is constrained, and what you can trust under pressure.

• Desired state is written into etcd; reality is observed through kubelet and controllers.
• Reconciliation closes the gap—until backpressure, admission, or latency distort the loop.
• Labels and selectors are not decoration; they are routing and ownership.

A minimal inspection ritual

Before you change anything, prove what the cluster believes and what the node is doing.

kubectl get ns
kubectl get nodes -o wide
kubectl get deploy,rs,pods -A -o wide
kubectl get events -A --sort-by=.lastTimestamp | tail -n 40

Common mistakes the Order sees

These failures present as “mystery outages”. They are usually foundations.

• Treating status as truth without reading conditions and events.
• Changing labels casually and severing Services or controller ownership.
• Confusing readiness (traffic gate) with liveness (kill switch).
• Assuming a YAML apply is instantaneous and consistent across controllers.