CRDs, Operators and API Evolution

Kubernetes' built-in object types (Pod, Deployment, Service, and so on) cover general workloads well, but teams often want the API to understand domain-specific concepts too, like "a PostgreSQL cluster" or "a TLS certificate." A CustomResourceDefinition (CRD) is what registers a brand-new object type with the API server, giving you a schema and validation for it - but a CRD alone is just inert data storage, exactly like an Ingress object without a controller. The behavior comes from a separate piece of software called an operator: a custom controller, following the same reconciliation-loop pattern from the foundations module, that watches instances of the new custom resource and does real work to make reality match them - provisioning a database cluster, or requesting and renewing a TLS certificate, for example. When something goes wrong with a custom resource, resist the temptation to treat it as an unknowable black box: inspect its schema with kubectl explain, read its status.conditions exactly like you would for a built-in object, and check the operator's own controller logs - the debugging method is identical, it is only the object type that is unfamiliar. Because CRDs and their schemas can change or be removed between Kubernetes versions, use API deprecation and discovery tooling to catch incompatibilities before a cluster upgrade, not during one.

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 manifest references an API version that has since been formally removed by the Kubernetes release the cluster is about to be upgraded to. Objects already stored using that old API version can keep working for a surprisingly long time - sometimes until the next restart, rollout, or the upgrade itself - which is exactly what makes this so dangerous: the incompatibility hides quietly until the worst possible moment. Reviewing which API versions are actually in use, and migrating off deprecated ones, belongs firmly in release and upgrade planning, not in the upgrade's own post-mortem.

Operate custom APIs like products

Use kubectl explain KIND.spec --api-version=GROUP/VERSION, inspect CRD schemas and printer columns, and read custom-resource conditions before controller logs. A CRD stores data; the operator supplies reconciliation. Version conversion, defaulting and validation determine whether old and new clients can coexist. Back up both custom resources and any external state the operator manages.

Scenario: An operator Pod is healthy but every resource reports a stale observedGeneration. Its watch permission was removed, so it cannot see new intent. RBAC and controller work queues are part of custom-resource health.
Tip: Before uninstalling an operator, inspect finalizers and deletion procedures. Removing the controller first can leave resources permanently Terminating and external infrastructure orphaned.

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 cert-manager-issuer-secret. Open /labs/kubernetes, choose cert-manager-issuer-secret, 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.