Exam Bundle

The API listing shows Pods associated with namespaces and Nodes listed as cluster-wide objects. That observed result would support another specific correction: Distinguish namespaced resources such as Pods from cluster-scoped objects such as Nodes and PersistentVolumes. The other scenario concerns: An auditor believes every object in a cluster belongs to a namespace, including Nodes and PersistentVolumes. The current Timberline Foods problem is: An engineer expects kube-scheduler to restart crashed containers and kubelet to choose initial node placement. The evidence required here is: A walkthrough shows the scheduler binding a Pod to a node and that node's kubelet reporting Pod lifecycle status.

A controlled test reports unscheduled Pods when capacity is exhausted and succeeds after approved resources are added. That observed result would support another specific correction: Explain that cluster capacity, scheduling constraints, quotas, and node availability bound achievable workload placement. The other scenario concerns: A team assumes a Kubernetes cluster can always run unlimited Pods regardless of node CPU, memory, or workload policy. The current Timberline Foods problem is: An engineer expects kube-scheduler to restart crashed containers and kubelet to choose initial node placement. The evidence required here is: A walkthrough shows the scheduler binding a Pod to a node and that node's kubelet reporting Pod lifecycle status.

An image is pulled from an approved registry and started by the node runtime without running inside the registry itself. That observed result would support another specific correction: Distinguish a registry that distributes image artifacts from the container runtime on a worker node. The other scenario concerns: A developer confuses an image registry with the place where Kubernetes runs its containers. The current Timberline Foods problem is: An engineer expects kube-scheduler to restart crashed containers and kubelet to choose initial node placement. The evidence required here is: A walkthrough shows the scheduler binding a Pod to a node and that node's kubelet reporting Pod lifecycle status.

Separate runtime configuration from the image through a ConfigMap mounted or exposed by the Pod specification. That choice addresses a different foundational issue: A team hard-codes feature toggles into container images so a small configuration edit triggers a full rebuild. The issue at Juniper Media is instead: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. It does not establish the required correction; the decisive verification would be: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

Distinguish the Secret API resource from storage encryption and require explicit encryption-at-rest and RBAC controls. That choice addresses a different foundational issue: A developer uses a Secret as proof that credentials are encrypted by default inside the cluster's backing store. The issue at Juniper Media is instead: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. It does not establish the required correction; the decisive verification would be: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

Distinguish resource spec as desired state from status fields reporting observed conditions and lifecycle phases. That choice addresses a different foundational issue: A team mixes desired configuration with observed health and assumes a Pending Pod is in the desired state because the manifest was accepted. The issue at Juniper Media is instead: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. It does not establish the required correction; the decisive verification would be: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

The resource ownership view shows Deployment to ReplicaSet to Pods and a replica count converging after a Pod is removed. That observed result would support another specific correction: Explain that a Deployment manages ReplicaSets, which in turn maintain the desired number of Pods. The other scenario concerns: A team thinks a Deployment directly stores every Pod's status and does not understand why ReplicaSets appear in its resource list. The current Juniper Media problem is: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. The evidence required here is: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

The Deployment remains listed with desired replicas zero and returns Pods when the replica target increases again. That observed result would support another specific correction: Explain that a zero replica target leaves the Deployment resource present while it maintains no Pods. The other scenario concerns: A team expects scaling a Deployment to zero to delete its manifest and forget that the controller still exists. The current Juniper Media problem is: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. The evidence required here is: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

The manifest records the requested Pod configuration while status reveals that scheduling is still pending. That observed result would support another specific correction: Distinguish resource spec as desired state from status fields reporting observed conditions and lifecycle phases. The other scenario concerns: A team mixes desired configuration with observed health and assumes a Pending Pod is in the desired state because the manifest was accepted. The current Juniper Media problem is: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. The evidence required here is: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

Explain that a reusable image holds application software while controller configuration declares how many Pod replicas are desired. That choice addresses a different foundational issue: A cluster design assumes the container image determines how many Pods a Deployment will run. The issue at Meadowbrook Finance is instead: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. It does not establish the required correction; the decisive verification would be: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

Match apiVersion and kind to the intended resource schema and place controller configuration under its proper specification. That choice addresses a different foundational issue: A developer copies a Pod manifest with the wrong kind field and cannot explain why the API rejects Deployment-specific properties. The issue at Meadowbrook Finance is instead: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. It does not establish the required correction; the decisive verification would be: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

Distinguish namespaced resources such as Pods from cluster-scoped objects such as Nodes and PersistentVolumes. That choice addresses a different foundational issue: An auditor believes every object in a cluster belongs to a namespace, including Nodes and PersistentVolumes. The issue at Meadowbrook Finance is instead: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. It does not establish the required correction; the decisive verification would be: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

A trainee can locate the control plane and worker nodes and identify the Pods running the sample service. That observed result would support another specific correction: Explain that a cluster combines a control plane with worker nodes running containerized workloads grouped into Pods. The other scenario concerns: An onboarding guide describes a Kubernetes cluster as one large application container rather than a group of managed machines and workloads. The current Meadowbrook Finance problem is: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. The evidence required here is: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

A validated manifest uses the correct resource apiVersion independently of its container image tag. That observed result would support another specific correction: Explain that apiVersion selects the Kubernetes resource API schema, whereas image tags identify application artifacts. The other scenario concerns: A tutorial equates the Kubernetes API version field with the container image's application release version. The current Meadowbrook Finance problem is: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. The evidence required here is: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

Resources with the same name can exist in distinct namespaces, while an explicit access-control test still enforces separation. That observed result would support another specific correction: Explain that namespaces organize namespaced resources and access policies, but do not by themselves provide strong multitenant isolation. The other scenario concerns: A team confuses a Kubernetes namespace with a virtual machine and assumes it supplies a fully independent security boundary by itself. The current Meadowbrook Finance problem is: An operator thinks kube-controller-manager is a dashboard server rather than a collection of reconciliation controllers. The evidence required here is: When replicas drift, controller activity restores the desired count without a dashboard running manual repair.

Distinguish namespaced resources such as Pods from cluster-scoped objects such as Nodes and PersistentVolumes. That choice addresses a different foundational issue: An auditor believes every object in a cluster belongs to a namespace, including Nodes and PersistentVolumes. The issue at Eastshore Energy is instead: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. It does not establish the required correction; the decisive verification would be: A request from another authorized API client observes the same persisted resource state as kubectl.

Use a Service or other stable discovery mechanism rather than relying on replaceable Pod names and IP addresses. That choice addresses a different foundational issue: A user believes the pod name is a stable identity that a client should store as its permanent service endpoint. The issue at Eastshore Energy is instead: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. It does not establish the required correction; the decisive verification would be: A request from another authorized API client observes the same persisted resource state as kubectl.

Describe a Pod as one or more tightly coupled containers sharing a network namespace and selected storage volumes. That choice addresses a different foundational issue: A development team calls each individual container a Pod and cannot explain why a logging sidecar shares networking with its main process. The issue at Eastshore Energy is instead: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. It does not establish the required correction; the decisive verification would be: A request from another authorized API client observes the same persisted resource state as kubectl.

The architecture review assigns cluster state to etcd, scheduling to kube-scheduler, and node execution to kubelet. That observed result would support another specific correction: Identify etcd as the key-value store for Kubernetes cluster state, not the workload runtime or scheduler. The other scenario concerns: A security reviewer thinks etcd is the scheduling daemon that starts user containers on worker nodes. The current Eastshore Energy problem is: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. The evidence required here is: A request from another authorized API client observes the same persisted resource state as kubectl.

The demo configuration is in a ConfigMap while secret values are separated and exposed only to authorized workloads. That observed result would support another specific correction: Use ConfigMaps for non-secret configuration and Kubernetes Secrets with suitable protections for credentials. The other scenario concerns: An application guide tells engineers to place database credentials directly inside a ConfigMap because it stores configuration. The current Eastshore Energy problem is: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. The evidence required here is: A request from another authorized API client observes the same persisted resource state as kubectl.

A permitted kubectl request reads the target resource without anyone logging into a control-plane host. That observed result would support another specific correction: Use the Kubernetes API through authorized clients such as kubectl and describe the API server as the control-plane entry point. The other scenario concerns: A novice believes Kubernetes API objects are changed only by SSH access to the controller machine. The current Eastshore Energy problem is: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. The evidence required here is: A request from another authorized API client observes the same persisted resource state as kubectl.

Define a Service as a stable logical endpoint selecting backend Pods, with cluster networking directing traffic to eligible endpoints. That choice addresses a different foundational issue: A team expects a Service to be a container process and cannot explain how clients reach changing backend Pod addresses. The issue at Alpine Healthcare is instead: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. It does not establish the required correction; the decisive verification would be: A request from another authorized API client observes the same persisted resource state as kubectl.

Separate runtime configuration from the image through a ConfigMap mounted or exposed by the Pod specification. That choice addresses a different foundational issue: A team hard-codes feature toggles into container images so a small configuration edit triggers a full rebuild. The issue at Alpine Healthcare is instead: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. It does not establish the required correction; the decisive verification would be: A request from another authorized API client observes the same persisted resource state as kubectl.

Use labels as key-value selectors for organizing resources and connecting Service or controller selection to Pods. That choice addresses a different foundational issue: A training lab treats YAML metadata labels as human comments and fails to use them for workload grouping. The issue at Alpine Healthcare is instead: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. It does not establish the required correction; the decisive verification would be: A request from another authorized API client observes the same persisted resource state as kubectl.

A Service selects only Pods with the matching label, while descriptive annotations remain non-selecting metadata. That observed result would support another specific correction: Use labels as key-value selectors for organizing resources and connecting Service or controller selection to Pods. The other scenario concerns: A training lab treats YAML metadata labels as human comments and fails to use them for workload grouping. The current Alpine Healthcare problem is: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. The evidence required here is: A request from another authorized API client observes the same persisted resource state as kubectl.

The Deployment remains listed with desired replicas zero and returns Pods when the replica target increases again. That observed result would support another specific correction: Explain that a zero replica target leaves the Deployment resource present while it maintains no Pods. The other scenario concerns: A team expects scaling a Deployment to zero to delete its manifest and forget that the controller still exists. The current Alpine Healthcare problem is: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. The evidence required here is: A request from another authorized API client observes the same persisted resource state as kubectl.

The demo configuration is in a ConfigMap while secret values are separated and exposed only to authorized workloads. That observed result would support another specific correction: Use ConfigMaps for non-secret configuration and Kubernetes Secrets with suitable protections for credentials. The other scenario concerns: An application guide tells engineers to place database credentials directly inside a ConfigMap because it stores configuration. The current Alpine Healthcare problem is: An internal reference chart treats kubectl as the authoritative storage system and ignores the Kubernetes API. The evidence required here is: A request from another authorized API client observes the same persisted resource state as kubectl.

Distinguish control-plane reconciliation and API decisions from the kubelet and runtime that execute Pods on worker nodes. That choice addresses a different foundational issue: A new platform operator assumes the control plane directly executes all application containers rather than coordinating the nodes. The issue at Brookside Manufacturing is instead: A developer confuses an image registry with the place where Kubernetes runs its containers. It does not establish the required correction; the decisive verification would be: An image is pulled from an approved registry and started by the node runtime without running inside the registry itself.

Explain that a reusable image holds application software while controller configuration declares how many Pod replicas are desired. That choice addresses a different foundational issue: A cluster design assumes the container image determines how many Pods a Deployment will run. The issue at Brookside Manufacturing is instead: A developer confuses an image registry with the place where Kubernetes runs its containers. It does not establish the required correction; the decisive verification would be: An image is pulled from an approved registry and started by the node runtime without running inside the registry itself.

Distinguish annotations as non-identifying metadata from labels used by selectors and other Kubernetes mechanisms. That choice addresses a different foundational issue: An engineer changes a descriptive annotation expecting it to route production traffic to a different set of Pods. The issue at Brookside Manufacturing is instead: A developer confuses an image registry with the place where Kubernetes runs its containers. It does not establish the required correction; the decisive verification would be: An image is pulled from an approved registry and started by the node runtime without running inside the registry itself.

The resource ownership view shows Deployment to ReplicaSet to Pods and a replica count converging after a Pod is removed. That observed result would support another specific correction: Explain that a Deployment manages ReplicaSets, which in turn maintain the desired number of Pods. The other scenario concerns: A team thinks a Deployment directly stores every Pod's status and does not understand why ReplicaSets appear in its resource list. The current Brookside Manufacturing problem is: A developer confuses an image registry with the place where Kubernetes runs its containers. The evidence required here is: An image is pulled from an approved registry and started by the node runtime without running inside the registry itself.

Resources with the same name can exist in distinct namespaces, while an explicit access-control test still enforces separation. That observed result would support another specific correction: Explain that namespaces organize namespaced resources and access policies, but do not by themselves provide strong multitenant isolation. The other scenario concerns: A team confuses a Kubernetes namespace with a virtual machine and assumes it supplies a fully independent security boundary by itself. The current Brookside Manufacturing problem is: A developer confuses an image registry with the place where Kubernetes runs its containers. The evidence required here is: An image is pulled from an approved registry and started by the node runtime without running inside the registry itself.

The deployment uses the same image with different approved configuration values across test and production. That observed result would support another specific correction: Separate runtime configuration from the image through a ConfigMap mounted or exposed by the Pod specification. The other scenario concerns: A team hard-codes feature toggles into container images so a small configuration edit triggers a full rebuild. The current Brookside Manufacturing problem is: A developer confuses an image registry with the place where Kubernetes runs its containers. The evidence required here is: An image is pulled from an approved registry and started by the node runtime without running inside the registry itself.