A Layered Troubleshooting Workflow

When something breaks in Kubernetes, jumping straight to a fix without a systematic approach usually wastes time and sometimes destroys the evidence you needed. A reliable workflow starts by explicitly naming the scope of the problem (one container? one Pod? one node? the whole namespace or cluster?) and what changed recently, since most incidents trace back to a recent change. From there, move outward through layers of evidence in order: the object's own status and spec, its recent events, individual container states and their reasons, application logs, whether its Service has ready endpoints, basic networking, resource usage against requests/limits, and finally node health if nothing narrower explains it. Three commands do most of this work, each for a different purpose: kubectl get gives you quick breadth across many objects at once; kubectl describe is richer, merging one object's full details with its related recent events in a single view; and kubectl logs gives you actual evidence from inside the running process itself. Two habits protect the evidence you are gathering: always preserve timestamps as you go (so a later timeline can be reconstructed), and change only one variable at a time - if you adjust two things simultaneously and the symptom changes, you will never know which one actually mattered.

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 responder, eager to "just fix it," deletes a Pod that is stuck failing before ever reading its previous container logs or its recent events. The owning controller (a Deployment or ReplicaSet, from earlier modules) faithfully creates a brand-new replacement Pod within seconds - so the visible symptom disappears immediately, which feels like progress - but the actual evidence of what caused the original failure is now gone forever along with the deleted Pod, and the same root cause is very likely to recur on the new one. The rule to internalize: observe and gather evidence first, and only mutate the state of the system once you have a genuine hypothesis to test.

Preserve evidence before repair

Begin with scope: one container, one Pod, one node, one namespace or the cluster. Record recent changes and time. Run broad get queries, describe the failing owner and child, order events, capture current and previous logs, then inspect endpoints, resources and nodes. A timeline separates trigger from controller reactions. Make the smallest reversible change only after a testable hypothesis exists.

Scenario: Deleting a CrashLooping Pod removes its previous logs and creates a new name while the bad Deployment template remains. The visual symptom briefly disappears and no cause changes.
Tip: Save command output with UTC timestamps and object UIDs. Names can be reused; UIDs and generations identify the exact incident objects.

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 pod-crashloop-debug. Open /labs/kubernetes, choose pod-crashloop-debug, 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.