Capacity Planning and Quotas

Two objects work together to keep a shared cluster or namespace from being consumed unfairly by one team or workload. A ResourceQuota sets hard caps on total consumption within one namespace - total CPU/memory requested, total object counts of a given type, and similar aggregate limits. A LimitRange operates one level down, at the level of an individual Pod or container, supplying sensible default requests/limits when none are specified, and optionally enforcing minimum/maximum bounds on what any single object may declare. Overcommitting a cluster's actual physical capacity relative to the sum of everyone's requests can improve overall utilization (since not everyone uses their full request simultaneously), but it directly trades that efficiency for a real risk of contention if usage spikes align badly across tenants. A frequently overlooked planning mistake is sizing a namespace's quota to exactly cover steady-state usage with zero slack - leaving no reserved headroom for the temporary extra Pods a rolling update needs (maxSurge, from the Deployments module), for replacing capacity lost during a failure, or for essential system Pods that also consume real resources on every node. The soundest way to think about capacity is as one combined equation, not several separate numbers: PodDisruptionBudgets, rollout surge, HPA's maximum replica count, and actual node capacity all draw from the same finite pool at the same time.

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's ResourceQuota is set to precisely match twelve replicas' steady-state resource requests, with zero spare room budgeted in. When a routine rolling update needs just one additional, temporarily-surging Pod to replace an old one safely (as described in the release-strategies module), that extra Pod's creation is rejected outright by the quota - and because no new Pod can ever become ready to take over, the old Pod it was meant to replace can never safely be removed either, deadlocking the whole rollout. Release capacity - the temporary headroom a safe deployment strategy genuinely requires - has to be explicitly budgeted into quota planning from the start, not assumed to be a rounding error.

Treat capacity controls as one equation

ResourceQuota caps namespace totals and object counts; LimitRange validates or defaults individual requests and limits. Deployment surge, HPA maximum, DaemonSet overhead, system reservations and PDB availability all consume headroom. Model loss of the largest node or zone plus a normal rollout, not only steady state. Track Pending reasons and quota rejections as capacity signals.

Scenario: A namespace exactly fits twelve steady replicas. A rollout with maxSurge one fails admission, so no new ready Pod exists to replace an old one. Reserve explicit release capacity or use a reviewed no-surge strategy with availability tradeoffs.
Tip: Report both physical utilization and requested allocation by failure domain. Either can become the binding constraint first.

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 rolling-update-hpa-scaling. Open /labs/kubernetes, choose rolling-update-hpa-scaling, 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.