Init, Sidecar and Adapter Patterns

Not every helper container belongs permanently inside the main application's Pod - Kubernetes recognizes a few distinct, well-understood patterns for combining containers, and picking the right one matters. Init containers run once, in strict order, before the main application starts, and each one must fully complete (exit successfully) before the next runs or the application starts - they are the right place for one-time prerequisites like "wait for the database to accept connections" or "run a migration." Sidecars, in contrast, run continuously alongside the application for its entire life, doing an ongoing job such as shipping logs to a central system or running a lightweight local network proxy. Two more specialized variants exist: an adapter container normalizes or reformats the main application's output into a shape some other system expects, while an ambassador container proxies connections to an external dependency on the application's behalf, hiding networking details from it. Recent versions of Kubernetes added native "sidecar" semantics (an init container marked to keep running) that improve startup and shutdown ordering guarantees. Regardless of which pattern you reach for, always ask the same architectural question: does this helper genuinely share the main application's lifecycle and scaling boundary, or is it secretly an independent service that only looks convenient bundled in?

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 database-migration init container is added to a Deployment's Pod template so that "the schema is always up to date." Because a Deployment can run many replicas, that init container now runs once per replica, in parallel, and every one of them tries to run the same migration against the same database at the same time. Whether that is safe depends entirely on whether the migration tool is idempotent and safe to run concurrently (or uses a leader-election lock) - the YAML syntax for an init container was never the hard part of this design.

Design helpers for failure

Init containers run sequentially and retry according to the Pod restart policy; they should be bounded and idempotent. A sidecar should expose health independently so its failure does not masquerade as application health. Allocate resources per container because an unbounded helper can evict or throttle its primary. When debugging a distroless image, use kubectl debug -it POD --image=nicolaka/netshoot --target=APP where policy permits rather than installing tools into production.

Scenario: An init container waits forever for DNS with no timeout. The Pod remains Init and never exposes application logs. Add bounded retries with actionable output and inspect kubectl logs POD -c INIT.
Warning: Database schema migration in every replica couples rollout concurrency to data correctness. Prefer a controlled, observable migration job when only one writer is safe.

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.