Events, Metrics and Conditions

Four different signal types are commonly available when diagnosing a Kubernetes problem, and knowing what each one is actually good for prevents wasted time staring at the wrong one. Events are short-lived, human-readable notes controllers emit about what they are doing ("FailedScheduling," "Pulled image") - genuinely useful during and shortly after an incident, but they are not a durable audit log and expire from the cluster after a limited retention window, by default just a few hours. Conditions are a controller's own structured judgment about an object's state, like Available or Progressing on a Deployment - more durable and precise than a freeform event message. Metrics (numeric time series like CPU usage, request latency, or queue depth) reveal resource saturation and quantitative application symptoms over time, while logs explain the specific discrete decisions and errors behind those numbers, and traces (where available) follow one individual request's full path through multiple services. None of these four alone tells the whole story - real diagnosis comes from correlating them together by timestamp, by the specific Deployment revision involved, and by Kubernetes events from the same window, building one coherent timeline rather than reading each signal in isolation.

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: CPU utilization graphs look completely normal across every replica, giving a false sense that resources are not the problem - yet incoming requests are visibly queuing up and response times are climbing. The real bottleneck turns out to be a fixed-size downstream connection pool (like a database connection pool) that has become fully exhausted, which produces almost no CPU signature at all, because waiting for a connection slot barely uses any processor time. Pure infrastructure metrics like CPU and memory are not enough on their own - genuine diagnosis needs application-level metrics too, such as request latency, saturation of internal resources like connection pools, and, for the distributed-systems labs specifically, protocol-level signals like Kafka ISR and consumer lag introduced in the StatefulSets module.

Correlate signals around one hypothesis

Metrics quantify rate, errors, duration and saturation; logs explain discrete decisions; traces follow request paths; events expose controller and node actions; conditions summarize reconciliation. Align them by timestamp, workload revision, Pod UID and node. kubectl top is convenient but limited historical evidence, so production diagnosis needs retained telemetry.

Scenario: HTTP latency begins immediately after a rollout, CPU stays low, connection-pool saturation rises and traces wait on a database span. Raising CPU limits cannot repair a dependency pool bottleneck.
Warning: Kubernetes events are aggregated and expire. Export them if incident reconstruction or audit requirements exceed their short retention, and never treat event absence as proof an action did not occur.

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 kafka-isr-consumer-lag. Open /labs/kubernetes, choose kafka-isr-consumer-lag, 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.