3V0-24.25 Advanced VMware Cloud Foundation 9.0 vSphere Kubernetes Service Questions and Answers

VCF 9.0 explicitly states: “The VKS exposes three layers of controllers to manage the lifecycle of a VKS cluster,” and then enumerates those layers. The first layer is the set of components that integrate the workload cluster with Supervisor-backed resources, including aCloud Provider Plug-inthat integrates with the Supervisor and enables infrastructure integrations such as persistent volume requests being passed to the Supervisor (which is integrated with Cloud Native Storage). The second layer isCluster API, described as providing “declarative, Kubernetes-style APIs for cluster creation, configuration, and management,” driven by resources that represent the cluster, the VMs making up the cluster, and cluster add-ons. The third layer is theVirtual Machine Service, which provides a declarative API for managing VMs and associated vSphere resources and is used to manage the lifecycle of the control plane and worker node VMs hosting a VKS cluster.

Therefore, optionAis the only answer that matches the three lifecycle controller layers defined in the VCF 9.0 documentation.

In Kubernetes, the component that actuallycreates and runs workloads on a nodeis thekubelet. The kubelet is the node agent that ensures the containers described by PodSpecs are running on that node. VCF 9.0 maps this concept directly into vSphere Supervisor by describingSphereletas “a kubelet that is ported natively to ESXi and allows the ESXi host to become part of the Kubernetes cluster,” showing that kubelet functionality is responsible for running workloads on worker nodes (ESXi hosts in the Supervisor case).

The other options have different roles:etcdis the control plane data store,API Serveris the front-end for Kubernetes API operations, and theSchedulerdecides placement (which node should run a pod). VCF 9.0 even calls out that “the Kubernetes scheduler… cannot place pods intelligently” without visibility into vCenter inventory—reinforcing that scheduling is about placement decisions, not the act of creating/running the workload on the node.

So, while the scheduler selects where a pod should run, thekubeletis the component responsible for actually instantiating and maintaining the workload on the target node.

VCF 9.0 explicitly documents thatVKS supports two CNI options: Antrea and Calico, and that thesystem-defined default CNI is Antrea. This directly eliminates Flannel and Cilium as default supported options for VKS clusters on Supervisor in this context. VCF 9.0 also describes how a vSphere administrator can view or change this setting in the vSphere Client underSupervisor Management → Configure → Kubernetes Service → Default CNI, further reinforcing that Antrea is the baseline/default choice.

From a policy perspective in the question, the requirement is Kubernetes-layer observability and control of pod communications “without additional licensing or overlay complexity.” Antrea is presented in VCF 9.0 as the default CNI and is implemented usingOpen vSwitch, with networking and network policy capabilities provided at the Kubernetes layer for pods and services. Because it is the documented default (and supported) option for new VKS clusters, selectingAntreabest aligns with the “default supported option” requirement.

Harbor is used as aprivate registry serviceto store and distribute container artifacts for Kubernetes consumption, which is exactly what’s needed for a multi-tenant platform where multiple teams require isolated repositories. The VMware documentation treats Harbor as aVMware Tanzu Harbor Registry service, including governance around who can operate it and how teams are separated intoprojects(a key multi-tenancy boundary). For example, vSphere privileges explicitly cover the ability tocreate or delete a Harbor registryand tocreate, delete, or purge Harbor registry projects, reinforcing that Harbor is operated as a managed registry with project-scoped administration and access control.

In practice for regulated environments, the registry role is not just storage—Harbor is commonly used to enforce enterprise controls likepolicy-driven access (RBAC), and it supports security capabilities such asimage vulnerability scanningandimage trust/signing, which directly address the requirement to prevent unsafe images from being promoted or deployed.

VCF 9.0 documentation explicitly indicates thatvCenter upgrades and the Supervisor/cluster (Workload Management) upgrade are distinct, noting that “if you have only upgraded vCenter and not the cluster” then DevOps engineers have reduced permissions until the cluster is upgraded. This supports the stated standard that VKS/Workload Management lifecycle can be treated separately from vCenter. For the private registry requirement, VCF 9.0 provides an operational mechanism to authenticate and pull artifacts from private registries: “Registry secrets allow package and repository consumers to authenticate to and pull images from private registries,” implemented via a standard Kubernetes Secret of type kubernetes.io/dockerconfigjson.

Taken together, the standard implies (1)asynchronoushandling (separate lifecycle from vCenter) and (2)privatesourcing (images pulled from an internal registry with registry secrets). Therefore, selectingAsynchronous Privatebest matches both requirements in a single configuration choice, aligning with the documented separation of upgrades and the documented need to use authenticated access to private registries.

Answer (Correct Order):

Select the target cluster in the workload domain.

Open Workload Management and select “Enable Supervisor Cluster”.

Choose cluster networking mode and stack.

Configure the control plane compute, networking, and storage policies.

Configure the workload network configurations.

Review, Validate, and Deploy Supervisor.

You start by selecting theexact vSphere cluster(in the workload domain) that will host the Supervisor, because Supervisor enablement is performedagainst a specific cluster. From there you launch the enablement workflow inWorkload Management(“Enable Supervisor Cluster”). Early in the wizard you must decide thenetworking mode(for example, VDS-based vs NSX-backed) and theIP stack, because those choices drive the remaining configuration screens and what inputs are required. Next you define theSupervisor control plane settings—compute sizing and the core policies the Supervisor will use (including storage policy selections and related defaults). After the control plane foundation is defined, you configure theworkload networkingused by namespaces and Kubernetes workloads (IP ranges, routing/LB integration depending on the selected mode). Finally, youreview/validateall inputs anddeployso the platform can create and configure the Supervisor control plane and supporting components.

VCF 9.0 states that you can provision Kubernetes clusters using both GUI and CLI approaches, and it explicitly calls out the CLI clients: “the VCF CLI and kubectl provide command-line interfaces for provisioning Kubernetes clusters.” That directly maps tokubectl (A)andVCF CLI (E)as supported clients for provisioning and lifecycle operations. Separately, VCF 9.0 explains that vSphere administrators can “manage and monitor vSphere Pods, VMs, and VKS clusters by using the vSphere Client,” which corresponds tovSphere UI (C)in the question. In addition, the vSphere Client is used to access Supervisor-facing self-service interfaces (for example, the Local Consumption Interface through the vSphere Client), reinforcing vSphere UI as an operational entry point for managing Supervisor-backed services and workloads.

By contrast,Cluster APIis a controller framework (not an operator “client” for admins in this context), andesxtop/esxcliare ESXi host tools that do not represent the documented, supported interfaces for provisioning and managing VKS clusters at the Kubernetes service layer.

VCF 9.0 describes VKS as providing self-service, declarative lifecycle management for Kubernetes clusters and explicitly identifiesCluster APIas the mechanism that enables Kubernetes-style lifecycle automation. The documentation states that “The Cluster API provides declarative, Kubernetes-style APIs for cluster creation, configuration, and management,” and that its inputs include resources describing “the virtual machines that make up the cluster” plus “cluster add-ons.” This directly maps to automated lifecycle management: a desired-state YAML is reconciled by controllers that create, update, and maintain the control plane and worker node resources without manual, imperative procedures.

The same VCF 9.0 content also notes VKS exposes “three layers of controllers” for lifecycle management and lists Cluster API as one of those layers, reinforcing that Cluster API is foundational for ongoing cluster operations (create, scale, upgrade, reconcile). Contour is an ingress controller (traffic routing), kubeadm is a Kubernetes bootstrap tool (not the VKS lifecycle controller framework), and Grafana is observability/visualization (not cluster lifecycle).

VMware Cloud Foundation 9.0 documents a dedicated backup-and-restore approach forWorkload Managementwhere different components use different tools. Forworkloadsrunning on vSphere Supervisor–based Kubernetes (bothvSphere PodsandVKS cluster workloads), the documented solution isVelero, specifically theVelero Plugin for vSphereinstalled and configured on the Supervisor. The “Considerations for Backing Up and Restoring Workload Management” table explicitly lists: “Backup and restore vSphere Pods — Velero Plugin for vSphere” and “Backup stateless and stateful workloads on a VKS cluster and restore to a cluster provisioned by VKS — Velero Plugin for vSphere.”

The same section also clarifies thatSupervisor backups(via vCenter file-based backup) are for restoring theSupervisor control plane/stateand VKS node VMs—not for restoring workloads themselves—so workloads must be backed up separately.

Therefore, the correct tool for backing up and restoring workloads on clusters provisioned by vSphere Supervisor isVelero (using the Velero Plugin for vSphere).

The problem describes demand spikes where capacity exists, but the application cannot meet demand—this typically indicates the cluster needs toscale out(more nodes/pods) automatically when load increases. In VCF 9.0, VKS supportsCluster Autoscaleras an optional package and specifically calls out improvements: “Cluster Autoscaler supports scaling from zero or to zero… You must have the autoscaler standard package installed.” This directly supports optionA(install cluster autoscaling) as the mechanism to dynamically add capacity during peak events and reduce it afterward, optimizing cost and operations while meeting bursts. A load balancer (including Foundation Load Balancer) helps distribute traffic acrossexistingendpoints, but it does not create new compute capacity when pods are pending due to insufficient nodes. Similarly, installing Contour relates to ingress (routing inbound traffic) and is not, by itself, a capacity scaling solution. Oversubscription is a risky workaround that can degrade performance and does not provide targeted, policy-driven elasticity. Therefore, enablingcluster autoscalingis the correct way to address burst demand when underlying resources are available.

VCF 9.0 explains pod fundamentals by describing how Workload Management introducesvSphere Pods, stating a vSphere Pod is “equivalent of a Kubernetes pod” and that it “runs one or more Linux containers.” This directly eliminates optionB, because a pod can includeone or morecontainers (not only one).

The vSphere 9.0 documentation further defines a KubernetesPodas “a group of one or more containerized applications that share such resources as storage and network,” and notes the containers inside a pod are “started, stopped, and replicated as a group.” That definition reflects Kubernetes’ scheduling and lifecycle model: Kubernetes treats the pod as the primary unit it places and manages together, which is why a pod is regarded as thesmallest deployable unitfor running containerized workloads in Kubernetes. OptionsCandDare incorrect because pods are Kubernetes objects (not “managed by Docker” as a smallest entity), and Kubernetes abstracts the underlying runtime/host so pods are not defined as being “deployed directly on the virtual machine” as a characteristic.

In VCF 9.0, VKS cluster provisioning and lifecycle management is built around Kubernetes-native, declarative APIs. The VMware documentation describes thatCluster API provides declarative, Kubernetes-style APIs for cluster creation, configuration, and management, and that its inputs include the cluster definition plus the resources describing the virtual machines and “cluster add-ons.” This is why Cluster API is the component associated withprovisioning(creating and managing) workload clusters in the VKS model.

By contrast,cert-manageris typically used to automate certificate issuance/renewal for in-cluster components, andCarvelis a set of tools often used to package, configure, and install Kubernetes software, but neither is the core provisioning controller for creating the clusters themselves. The VCF documentation is explicit that Cluster API is the API layer responsible for cluster creation/configuration/management, making it the correct answer for provisioning VKS workload clusters (including those intended to run service mesh capabilities).

VCF 9.0 describes backing up workloads on VKS clusters usingVelero(Velero Plugin for vSphere) and shows the backup command pattern scoping backup content bynamespace(for example, velero backup create ... --include-namespaces=my-namespace). This demonstrates thatNamespacesare a primary object type within backup scope (you choose which namespace content is captured). The same section explains that after backup execution,Kubernetes metadata is uploaded to the object store. “Kubernetes metadata” in this context includes standard namespaced resources that represent running workloads and how they are exposed, such asServices(a core Kubernetes API object used to provide stable access to pods).

In addition to namespace scoping, Velero’s selection model commonly includes filtering backed-up Kubernetes objects using selectors (for example, selecting subsets of resources), which aligns with the “objects identified by a label selector” option as an object-selection mechanism within the scope of what Velero captures from Kubernetes APIs (metadata/resources) once the namespace is targeted.

Therefore, the best match to the documented Velero workload-backup model is:Namespace,Service, andlabel-selected objectswithin the included namespace(s).

Istio is available in VCF 9.0 as an optional package that can be installed for VKS clusters. The VCF 9.0 documentation explicitly lists “Istio” among the optional packages that “can be optionally installed” for the vSphere Kubernetes Service (VKS). In Kubernetes platforms, Istio is a service mesh that secures and manages service-to-service (east-west) traffic by establishing authenticated and encrypted connections between workloads. The encryption mechanism it provides for inter-service communication ismutual TLS (mTLS), which means both ends (client and server workloads) authenticate each other and negotiate encrypted transport for every service call—meeting the requirement to “encrypt the transport and communications between services.” This is distinct from host-level mechanisms like IPsec/IKEv2 (network-layer) or “AES-256” (an algorithm, not the service-to-service transport model). Enabling Istio Service Mesh is therefore aligned to deliver encrypted, identity-aware service communications across the cluster at the application service layer.

Kubernetes Role-Based Access Control (RBAC) is implemented through theRBAC API group(rbac.authorization.k8s.io) and defines the core authorization primitives used to grant permissions to users, groups, and service accounts. The cluster-scoped objects declared by the RBAC API areClusterRoleandClusterRoleBinding. AClusterRoledefines a set of permissions (verbs such as get/list/watch/create/update/delete) over resources at thecluster scope(including cluster-wide resources and optionally namespaced resources across namespaces). AClusterRoleBindingthenbindsthat ClusterRole to a subject (user/group/serviceaccount), making those permissions effective cluster-wide.

This differs from namespace-scoped RBAC objects (RoleandRoleBinding) which apply only within a single namespace. The other options are incorrect becauseClusterObject/ClusterNodeare not RBAC API objects,ValidatingAdmissionPolicybelongs to the admission control API surface (policy enforcement),ResourceQuotais a namespace resource governance object, andContainer/Deploymentare workload/runtime concepts defined in the core/apps APIs rather than authorization primitives.