Episode 8. Kubernetes control plane. Day 8 of your 12 day sprint.

Scenario: "Design zero-downtime tenant moves across regions." Answer with the move runbook, not the YAML.

Say: "State sticks to identity. A StatefulSet binds each pod to its persistent volume, and the operator reconciles desired versus actual."

Say the safety trio: "PodDisruptionBudgets bound disruption. Readiness gates lag before serving. Liveness stays separate from readiness."

Say the move order: "Drain, fence the old writer, promote, verify lag and RPO, cut traffic, keep rollback ready."

Lab: draft a one-page tenant-move runbook with prechecks, move steps, verify queries, and a rollback that works at three in the morning.

Trap to rule out: Say: "Stateless rollout habits kill stateful data. I never treat a database pod like a web pod."

Drill: recite the move order: drain, fence, promote, verify, cut, rollback ready.
Say: readiness gates lag, liveness stays separate, budgets bound disruption.
Say the Nomad line: container GC history is why fencing gets its own paragraph.
Done-check: your one-page runbook has prechecks, verify queries, and rollback.
Depth tag tonight: Debug on the runbook, Recognize on operator internals.

Teaser: tomorrow, Episode 9, tail latency. Averages lie, and the histogram tells the truth.