Disruption Budgets and Maintenance

A PodDisruptionBudget (PDB) is a safety rail around planned, voluntary actions - things like a node drain during maintenance, or a cluster-autoscaler-initiated eviction - and it has no effect at all on unplanned failures like a crash, an OOM kill, or a node abruptly dying; those are simply outside what a PDB governs. You express a PDB using either minAvailable ("at least this many replicas must stay up during any voluntary disruption") or maxUnavailable ("at most this many may be taken down at once"), and the Kubernetes eviction API checks the PDB before allowing a voluntary eviction to proceed, refusing it if honoring the budget is impossible right now. It is important to internalize that a PDB is a constraint, not a source of resilience by itself - it can only ever protect redundancy that genuinely already exists. If a workload has only one replica and its PDB says minAvailable: 1, that is not extra safety, it is simply a permanent block on ever voluntarily evicting that single Pod at all.

Analogy: Kubernetes is a thermostat, not a remote control. You declare the temperature; independent controllers measure reality and keep acting. Debugging means finding which sensor, rule or actuator prevents convergence.

Read the object as evidence

kubectl get RESOURCE NAME -n NAMESPACE -o yaml
kubectl describe RESOURCE NAME -n NAMESPACE
kubectl get events -n NAMESPACE --sort-by=.metadata.creationTimestamp
kubectl explain RESOURCE.spec

Do not memorize these as a ritual. The first command exposes desired and observed state, the second connects conditions and events, the third supplies a timeline, and the fourth checks the server's schema. Compare what the controller was asked to do with what it reports doing. Then test the narrowest hypothesis.

Scenario: A node drain during routine maintenance hangs indefinitely, refusing to finish, because several workloads on that node have PodDisruptionBudgets that are mathematically impossible to satisfy in the current situation - for example, a workload already running below its declared minAvailable due to an earlier, unrelated failure. The correct response is to actually read the eviction error messages the drain command reports, and then deliberately decide - scale the workload up first, or adjust the budget consciously - rather than reaching straight for --disable-eviction or force-deleting Pods, which throws away the exact safety guarantee the PDB exists to provide.

Budget voluntary disruption realistically

A PDB controls eviction API decisions for voluntary actions such as drain; it does not prevent node failure, direct Pod deletion or every controller update. Check disruptionsAllowed, currentHealthy and desiredHealthy before maintenance. PDB selectors must match the intended Pods, and readiness determines healthy count. Combine budgets with topology spread and enough replicas across failure domains.

Scenario: A three-member quorum has maxUnavailable: 1, but two members already share a failed node. The PDB could not correct placement and cannot restore quorum. Availability architecture precedes disruption policy.
Warning: Forcing drain with deletion or disabled eviction bypasses the budget. That may be necessary for a dead node, but it is an explicit risk decision requiring workload-owner input.

Production reasoning

Ask four questions: Who owns this object? What dependency must become ready next? Which controller reports the blocking condition? What evidence would disprove my current theory? This prevents symptom-driven changes. Record the context, namespace, object generation, image digest and recent rollout before mutation; a recreated Pod may erase the evidence you needed.

Warning: Running is not the same as ready, healthy, durable or correct. Kubernetes status is layered. Confirm the application-level outcome as well as the object state.
Goal: Put this model into practice in the Kubernetes lab taints-pdb-maintenance. Open /labs/kubernetes, choose taints-pdb-maintenance, predict the failure path before changing anything, then use the simulator's check command to validate the finished state.

Deliberate practice

Before the lab, write the expected object relationship and the first three commands you will run. Afterward, explain why the fix converged and name one tempting change that would only mask the symptom. Repeat using an explicit namespace and a structured output format. This prediction-observation-explanation loop is what turns command familiarity into production judgment.

45-minute investigation

  1. Map (5 min): draw the owner-to-child chain and mark every namespace, selector, identity and dependency involved.
  2. Predict (5 min): write one expected status condition, one likely event and one log or metric signal before opening the lab.
  3. Observe (10 min): collect YAML, describe output and ordered events. Do not mutate state. Record which observation disproves your first theory.
  4. Repair (15 min): make the smallest declarative correction, watch the responsible controller converge, and verify the user-facing path rather than stopping at Running.
  5. Stress (10 min): change one relevant constraint - replica count, label, readiness, resource value or placement rule - predict the outcome, observe it, then restore the known-good declaration.
Tip: Keep a short incident note with symptom, evidence, hypothesis, change, result. Across four sections this produces a reusable runbook instead of a pile of remembered commands.