ServiceAccount Least Privilege

As introduced in the configuration module, every Pod authenticates to the API as a ServiceAccount, and the single most impactful RBAC habit you can build is giving each distinct workload its own purpose-built ServiceAccount, bound only to the specific permissions that workload actually needs - rather than everything sharing one broad, generic identity. It is worth being explicit that a namespace boundary is not, by itself, a security boundary here: nothing stops a RoleBinding or ClusterRoleBinding from granting a ServiceAccount permissions in a completely different namespace, or cluster-wide, if someone writes that binding - namespaces organize resources, they do not automatically wall off permissions unless RBAC is written to respect that boundary. For any workload that genuinely never calls the Kubernetes API itself, explicitly set automountServiceAccountToken: false so it is never handed a credential it has no use for and cannot leak. And prefer the modern pattern of short-lived, audience-bound projected tokens (which automatically rotate and expire) over the legacy pattern of a long-lived static Secret token, which if leaked remains valid indefinitely.

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: Every application in a namespace is initially deployed using that namespace's default ServiceAccount out of convenience, since it requires no extra YAML. Months later, one particular controller genuinely needs broad permissions, and rather than creating it a dedicated ServiceAccount, someone grants those broad permissions to the shared default ServiceAccount instead, because "it's already what everything uses." Now a security compromise in literally any application in that namespace - even a completely unrelated one that never needed elevated access - inherits that controller's full privileges, because they all share one identity. One workload, one ServiceAccount, one minimal set of bound permissions is what actually contains a compromise to the thing that was compromised.

Give workloads distinct identities

Create one ServiceAccount per trust boundary, set serviceAccountName, and disable token automount for workloads that never call the API. Projected bound tokens are short-lived and audience-scoped; applications must tolerate rotation. Bind controllers to the namespaces and resources they actually reconcile and avoid using the default ServiceAccount as a shared privilege bucket.

Scenario: A metrics sidecar is compromised and steals the Pod's token. Because the application and sidecar share one Pod identity, it inherits the controller permissions too. Separate highly privileged agents or redesign the trust boundary.
Tip: Audit API calls by ServiceAccount and verify unused permissions over time. Least privilege is maintained from evidence, not completed once at deployment.

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 rbac-least-privilege. Open /labs/kubernetes, choose rbac-least-privilege, 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.