NCP-AI Nutanix Certified Professional - Artificial Intelligence NCP-AI 6.10 Questions and Answers

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

Nutanix Enterprise AI container images and Helm charts are hosted in a private Nutanix repository. To access these resources during installation, the administrator must obtain a specific set of credentials (often referred to as a "Docker Registry" or "Image Pull" secret).

According to the installation guide, the administrator must log in to the Nutanix Support Portal, navigate to the Downloads section, and select Nutanix Enterprise AI (product=nai). On this page, there is a specific option to generate/view the "Docker Hub Access Token" or credentials required to pull the NAI images. These credentials are then used in the Helm install command (e.g., --set imagePullSecret.credentials.username=...) to authenticate the cluster against the Nutanix registry.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

According to the Nutanix Enterprise AI network requirements and installation guide, the administrator must configure a fully qualified domain name (FQDN) on the DNS domain that resolves to the external IP address of the Ingress Gateway service.1 In the architecture used by Nutanix Enterprise AI (which typically leverages the KServe stack), Istio provides the Ingress Gateway (often referred to as the Envoy Ingress Gateway or Istio Ingress Gateway).2

The documentation explicitly states: "Ensure that you configure a fully qualified domain name (FQDN) on the DNS domain that is accessible to your Kubernetes cluster using the external IP address of the Envoy Ingress Gateway service ." 3 Since Envoy acts as the data plane and ingress controller for the Istio service mesh in this deployment, the DNS record must point to the LoadBalancer IP allocated to the Istio/Envoy gateway to route traffic correctly to the AI endpoints.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

Based on the Nutanix Enterprise AI (NAI) role-based access control (RBAC) and model management documentation:

Manual Upload Permissions: The action shown in the exhibit corresponds to a Manual Import (or "Manual Upload") of a model using an infrastructure storage source (NFS File Share). This workflow is typically used for "Dark Site" (air-gapped) environments or custom model ingestion.

Role Restrictions: By default or in secure configurations, the AI/ML User role is often restricted from performing manual uploads from infrastructure sources like NFS shares or S3 buckets. This capability generally requires the Allow Manual Upload setting to be explicitly enabled by an admin, or the action must be performed by an AI/ML Admin .

Infrastructure Access: Connecting to a "File Share" via NFS (as shown with fs01.nutanix.edu) involves mounting infrastructure storage to the NAI cluster. Standard AI/ML Users typically do not have the underlying privileges to define these storage mounts due to security risks. They are primarily consumers of models or importers from public hubs (like Hugging Face) if allowed, rather than infrastructure architects.

Conclusion: The import failed because the user attempting the action (an AI/ML User) lacked the necessary permissions to execute a manual import from a backend file share.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

According to the Nutanix Enterprise AI documentation regarding "Viewing Sample API Code," the sample code provided by the dashboard is designed to help developers integrate the endpoint into their applications. The documentation explicitly states: "The sample code contains the URL and API key required to access the endpoint."

These two components are essential for the client application to locate the service (Endpoint URL) and authenticate the request (API Key). User credentials (such as username/password) are not part of the standard inference API request, and infrastructure details like GPU type are abstracted away from the client code.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

The scenario describes a situation where the infrastructure is seeing high utilization ("resource consumption issue") but application performance is stable (consumers are happy). The immediate task is verification and analysis, not blind scaling.

Troubleshooting Step: The administrator should access the NAI Dashboard and review the Cluster & Node Usage (specifically the Infrastructure Summary widget). 8

Goal: This allows the admin to validate the operations team's report by identifying exactly which resource (CPU, Memory, or GPU) is nearing exhaustion and correlating it with the "additional workloads" recently added. Only after confirming a legitimate bottleneck (exhaustion) should the admin proceed to "increase it" (add resources), making Option C the correct and complete troubleshooting workflow.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

In the Nutanix Enterprise AI (NAI) workflow, API keys are mechanism-specific credentials used to secure access to inference services.

Dependency: The documentation for creating API keys explicitly states that the key must be assigned to an Endpoint .

Workflow: When navigating to the "API Keys" section to generate a new key, the user is required to select an existing endpoint from a dropdown menu to associate the key with. Consequently, you cannot create a functional API key in a vacuum; you must Create an Endpoint first (which itself implies you have already imported a model) before you can generate the credentials to access it. The endpoint acts as the parent object that the key authenticates against.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

According to the Nutanix Enterprise AI (NAI) deployment and installation documentation, managing the helm repository is a critical step in the installation workflow. The process involves adding the Nutanix chart repository and then identifying the correct chart version to install.

Helm Search Command: The correct command to query a repository for specific chart information is helm search. To view all available versions of the NAI Core chart (rather than just the latest version), the --versions flag must be appended. Therefore, the syntax helm search repo ntnx-charts/nai-core --versions is the specific command used to list all available versions of the nai-core chart within the ntnx-charts repository.

Incorrect Options Analysis:

helm pull is used to download a chart to the local directory, not to search or list available versions in the remote repository.

helm repo list is used to list the repositories that have been added to the local helm client (e.g., verifying that ntnx-charts exists), but it does not list the content or versions of charts inside those repositories.

helm repo get is not a valid Helm command for retrieving chart versions.

This procedure ensures the administrator can select the specific target version required for the environment before executing the helm install command.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

Nutanix Enterprise AI (NAI) is explicitly architected to support and validate models from two primary ecosystem sources/formats: NVIDIA NIM (NVIDIA Inference Microservices) and Hugging Face.

The NAI "Import Model" interface allows administrators to select models directly from the Hugging Face Model Hub (supporting various model architectures hosted there) or from the NVIDIA NGC Catalog (importing NVIDIA NIMs). The documentation repeatedly refers to "Importing a Large Language Model from Hugging Face" and "Importing NVIDIA NIMs" as the core supported workflows. While other formats like ONNX exist, the official platform support for importing and packaging focuses on the NIM containers and Hugging Face repositories.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

In the Nutanix Enterprise AI user interface, the management of API keys is located within the Endpoints section. To deactivate a key, the administrator must navigate to the Endpoints page from the left navigation pane.

Once on the Endpoints page, there is a specific button or tab labeled API Keys . Clicking this opens the API Keys list view. The administrator can then select the specific key (via checkbox) and use the Actions dropdown menu to select Deactivate . This location is distinct from the "Settings" or "Admin Configuration" menus.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

To successfully configure and validate the integration, the administrator needs to utilize the connection details provided by the NAI platform.

Validation Workflow: The correct procedure is to retrieve the Endpoint URL and the unique API Key from the NAI Endpoint details page (often via the "View Sample Request" or "Sample Code" feature).

Configuration: The administrator or developer must then configure the application to send HTTP requests to this specific URL with the API Key included in the header (e.g., as a Bearer token).

Testing: Sending a test request using these exact credentials allows the administrator to confirm the response meets requirements (schema and content) and validates that the application can successfully authenticate and communicate with the inference service before production rollout.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

According to the Nutanix Enterprise AI installation prerequisites, the administrator must configure a specific storage class for the persistent volumes that store Large Language Models (LLMs) and model data. The documentation explicitly states: "Ensure that you create a ReadWriteMany (RWX) storage class that enables static or dynamic provisioning of NFS shares for persistent volumes."

This ReadWriteMany access mode is mandatory because the AI models stored on these volumes must be simultaneously accessible (readable and writable) by multiple pods running on different worker nodes across the Kubernetes cluster. This shared access ensures that as inference endpoints scale out to multiple replicas, all instances can mount and utilize the same underlying model files without locking issues.

**Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:**

Nutanix Enterprise AI (NAI) is a cloud-native application that requires a Kubernetes layer to operate. The supported Nutanix-native Kubernetes platform for deploying NAI is the **Nutanix Kubernetes Engine (NKE)** (often managed via the Nutanix Kubernetes Platform, NKP).

NKE provides the certified CNCF-compliant Kubernetes clusters running on top of Nutanix Cloud Infrastructure (NCI) that NAI requires to orchestrate containers and manage GPU resources. While NAI can effectively run on public cloud Kubernetes services (like EKS or AKS), within the context of Nutanix-provided platforms, **Nutanix NKE** is the designated supported deployment platform for the Kubernetes layer. (Note: On-premises NCI is the underlying infrastructure, not the AI deployment platform itself).

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

In the Nutanix Enterprise AI (NAI) endpoint lifecycle, the Pending status has a specific definition related to infrastructure resource availability. According to the endpoint management documentation:

" Pending : The system is waiting for the resources required to create the endpoint. For example, if the required number of GPUs are not available, Nutanix Enterprise AI maintains a Pending status until the GPUs become available."

The exhibit shows the administrator explicitly requested 1 NVIDIA-H100-NVL GPU for the instance. If the underlying Kubernetes cluster does not have a free H100 GPU (e.g., all H100s are currently occupied by other active endpoints or training jobs), the scheduler cannot place the pod, and the status will remain "Pending" indefinitely until a GPU is freed.

In contrast:

Unauthorized NGC Keys typically result in a Failed status due to image pull errors (ImagePullBackOff) or authentication failures, not a resource-wait "Pending" state.

Permissions issues would generally prevent the creation wizard from completing or result in an immediate 403 Forbidden error upon submission.

Wrong GPU selection (compatibility) is unlikely here as the "gemma-2-2b-it" is a small model that can easily run on the powerful H100; the issue is simply that the requested hardware is currently fully utilized.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

To investigate the removal of an API key, the administrator must use the Audit Events page in the Nutanix Enterprise AI dashboard. The audit logs capture specific actions performed on system entities.

Since the issue is a missing key, the specific event to look for is the deletion of that key. The correct filter configuration is:

Entity Type: API Key (This isolates the log to events involving keys, not endpoints or users).

Action: Delete (This specifically looks for the removal event).

This filter combination will display who performed the delete action and when it occurred.

Here are the verified answers and detailed explanations based on the Nutanix Enterprise AI (NAI) documentation.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

In the Nutanix Enterprise AI dashboard, the API Key management interface allows administrators to control the lifecycle of access tokens. The available actions for an API key are to "Activate" or "Deactivate" it.

According to the API Keys documentation, when an administrator selects an active key and clicks the Deactivate action, the system immediately revokes access for that key and updates its status column to Inactive . This state allows the key to be retained in the system (meeting the corporate policy of not deleting it) and permits the administrator to switch it back to "Active" status later if access needs to be restored. "Expired" and "Failed" are not status labels resulting from this specific manual deactivation action.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

In Nutanix Enterprise AI, API keys are securely managed and are strictly associated with the user who created them. For security reasons, once an API key is created and displayed, it is masked and cannot be viewed again in the dashboard.

If an AI/ML Admin attempts to test an endpoint and relies on an API key created by another AI/ML Admin, they will be unable to retrieve or view that key (as it is masked for all users except the creator at the moment of creation). Furthermore, if the client application is using a key that the current admin cannot validate or manage, access may fail if that key was revoked or not shared securely offline. The visibility and management of API keys are scoped to the creator; therefore, the testing admin should generate their own API key and assign it to the endpoint to ensure successful access.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

For text-generation models (LLMs), Nutanix Enterprise AI provides a built-in "Test" button in the dashboard that allows users to send sample or custom prompts directly from the UI. However, the documentation explicitly states that this internal Test functionality is not supported for non-text generation endpoints, such as Image Generation or Image Classification models.

To test these specific non-text endpoints, the administrator must access the Endpoint Access widget (or "View Sample Request" dialog), copy the Sample API Code (which includes the correct URL and necessary API key), and run it externally in a terminal (using curl) or within the client application. This manual execution of the sample code is the verified method for validating non-text endpoints.

Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:

The core management components of the Nutanix Enterprise AI application, including the User Interface (UI), the API server, and the database, are deployed into the nai-system namespace.

If the UI is failing to launch, the administrator must verify the status of the application pods specifically within this namespace. The command kubectl get pods -n nai-system will list the status of pods such as nai-ui, nai-api, and nai-db.

knative-serving contains the inference server components, not the management UI.

kube-system contains core Kubernetes components.

ntnx-base typically contains platform-level agents but not the NAI application logic.

**Comprehensive and Detailed Explanation From Exact Extract of Nutanix Artificial Intelligence documents knowledge without any external URL or links:**

To make a Large Language Model (LLM) available for deployment in Nutanix Enterprise AI, the administrator must use the **Models** section within the Nutanix Enterprise AI (NAI) dashboard. The correct workflow involves using the **Import Model** interface.

Within this interface, the administrator can choose to import a model directly from the **Hugging Face** model hub (by providing the repository name and optional token) or perform a **Manual Import** (Upload) for custom or air-gapped models. This action registers the LLM in the NAI repository, changing its status to "Ready," which is the prerequisite for deploying it as an inference endpoint. Options regarding Prism Central marketplace or third-party orchestration tools are incorrect for the specific task of ingesting the model *into* the NAI platform.