Build From Default Deny

The generally recommended way to introduce NetworkPolicy into an existing namespace is to start from default-deny for both ingress and egress (a policy selecting all Pods with empty allow rules for each direction), and then deliberately, incrementally permit only the specific flows you know are actually needed - DNS lookups, gateway-to-application traffic, and application-to-dependency traffic being the near-universal first three. When writing the peers a rule allows, you have three selector types to combine carefully: podSelector (Pods matching certain labels, within the policy's own namespace by default), namespaceSelector (any Pod within namespaces matching certain labels), and ipBlock (a raw CIDR range, typically reserved for traffic to/from outside the cluster entirely). Pay close attention to selectors written as empty objects, because their meaning is broader than it looks at a glance - an empty podSelector: {} in the policy's own subject field means "every Pod in this namespace," and an empty object inside a peer list can mean "allow from anywhere" - both are exactly the kind of thing that deserves a careful second reviewer before merging.

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 namespace has egress default-deny applied, and shortly afterward, every single application in it starts reporting failures connecting to its database - by hostname. The actual cause is that DNS lookups themselves are also outbound (egress) traffic, and the default-deny policy blocks the Pod's queries to CoreDNS just as thoroughly as it blocks the intended database connection - so hostname resolution itself never even succeeds, and every downstream connection attempt fails before ever reaching the network layer the team assumed was the problem. DNS egress needs to be explicitly permitted as one of the very first allow rules added, before debugging anything else.

Introduce default deny without losing dependencies

Apply ingress and egress isolation separately, inventory flows, and first permit DNS, monitoring and required control-plane or external endpoints. Build narrow allows using namespace labels plus Pod labels so a same-named Pod in an untrusted namespace is not admitted. Remember that podSelector: {} selects all Pods in the policy namespace, while an empty peer object can mean all sources.

Scenario: After egress isolation, applications report database hostname failures. DNS was omitted, so no connection ever reached the database. Validate resolution before debugging TCP policy.
Tip: Stage policies in a test namespace with the same labels, capture expected flows, and alert on denied traffic before enforcing broadly.

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 networkpolicy-kong-routing. Open /labs/kubernetes, choose networkpolicy-kong-routing, 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.