DNS, kube-proxy and Data Planes

CoreDNS is the cluster's internal DNS server, and it is what makes a Service reachable by a friendly, stable name such as api.team.svc.cluster.local (service-name.namespace.svc.cluster-domain) instead of a raw virtual IP. Pods are also configured with DNS search domains, which is why a short, unqualified name like api can resolve correctly from inside the same namespace - the resolver silently tries appending each search domain in turn until one succeeds. Resolving a name to an IP address is only half the journey, though; something still has to actually forward packets sent to that Service's virtual IP toward one of the real Pod IPs behind it. That job usually belongs to kube-proxy, which commonly programs the Linux kernel's iptables or IPVS packet-filtering rules to rewrite and route that traffic - though some modern CNIs replace kube-proxy entirely with an eBPF-based dataplane instead. A crucial, often-surprising fact: there is genuinely no process listening on a Service's ClusterIP the way a normal server listens on a socket; it is a purely virtual address that only exists as a set of forwarding rules.

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 application is moved into a new namespace and, out of habit, keeps referring to its dependency using the short, unqualified name redis instead of the fully namespace-qualified form. Thanks to DNS search-domain behavior, that short name now resolves to a completely different Service (one that happens to also be named redis, in the new default search namespace) - or to nothing at all - rather than the intended one, and it does so silently, with no error until requests start failing oddly. Use fully qualified Service names whenever a reference crosses a namespace boundary, and inspect a Pod's /etc/resolv.conf directly whenever search-domain behavior produces a surprising result.

Separate naming from forwarding

Check getent hosts api.team.svc.cluster.local or nslookup inside the failing Pod, then inspect /etc/resolv.conf and CoreDNS only if resolution fails. DNS success proves a name became an address, not that endpoints listen. Service forwarding may be implemented by iptables, IPVS or eBPF; the ClusterIP usually has no socket to find with ss. Inspect the implementation appropriate to the cluster.

Scenario: Intermittent five-second delays occur only for short names. A long search list and ndots cause multiple queries before the intended external name. Test the fully qualified name and observe DNS traffic.
Tip: Use namespace-qualified Service names in configuration crossing namespace boundaries; reserve short names for intentional same-namespace coupling.

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 service-dns-networking. Open /labs/kubernetes, choose service-dns-networking, 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.