Snapshots, Backups and Stateful Recovery

"Persistent" only means the data survives normal Kubernetes lifecycle events like a Pod restart - it is a much weaker guarantee than "backed up," and conflating the two is one of the most common and expensive mistakes in operating stateful workloads. A CSI snapshot captures the state of a volume's underlying blocks at one instant, and unless the application was made aware and told to pause or flush its writes first (a concept called quiescence), that snapshot may capture data mid-write - technically "crash-consistent" (as safe as an unexpected power loss) rather than cleanly consistent. A genuine backup strategy needs retention policy (how long to keep old backups), encryption, and, most importantly, restore tests performed on a schedule - a backup nobody has ever successfully restored from is only a hypothesis. It also helps to keep two different kinds of "recovery" mentally separate: recovering the Kubernetes objects themselves (a StatefulSet's YAML, its PVC definitions) is a completely different exercise from recovering the actual application data sitting inside those volumes, and a full incident-recovery plan has to document both, including exactly what happens to backing storage if a PVC or StatefulSet is deleted (recall reclaimPolicy from earlier).

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: An automated backup job for a PostgreSQL database reports success every single night for months, giving the team false confidence - but when a real restore is finally attempted after an incident, the restored data directory is internally inconsistent and the database refuses to start cleanly, because the backup was only ever a raw filesystem snapshot taken without coordinating with Postgres's own write-ahead log. Application-aware backup tooling (like pg_dump, or a snapshot explicitly coordinated with the database's own consistency mechanisms) exists precisely to avoid this trap, and the only way to know a backup strategy actually works is to routinely and deliberately prove it by performing a full test restore, not by watching a job status turn green.

Prove restore, not backup creation

Define recovery point and recovery time objectives, then choose application dumps, coordinated filesystem snapshots or storage replication accordingly. A CSI snapshot captures blocks; it may not capture a transactionally consistent database unless the application quiesces or can recover its journal. Keep backups outside the cluster and failure domain, encrypt them, and rehearse restoration into an isolated namespace.

Scenario: Nightly snapshots are green, but restore needs encryption keys stored only in the failed cluster. The data exists and is unusable. Recovery inventories must include credentials, operators, classes and external dependencies.
Tip: Validate row counts, checksums and application queries after restore. A Bound PVC and Running database Pod prove infrastructure assembly, not business-data correctness.

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 init-pgbouncer-postgres. Open /labs/kubernetes, choose init-pgbouncer-postgres, 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.