NCP-AI 시험 - Nutanix실제시험문제와 답 - 75문항

Question No : 1

Corporate policy says that API keys should be deactivated, not deleted, when a compromise is suspected, but may need to be restored.
Which new status will the key display after the one-click Deactivate action?

A.Pending
B.Inactive
C.Expired
D.Failed

정답:

Question No : 2

An administrator has been asked to install Nutanix Enterprise AI and is defining the NFS storage class for storing models.
What type of access mode needs to be defined?

A.ReadWriteOnce (RWO)
B.ReadWriteMany (RWX)
C.ReadOnlyMany (ROX)
D.ReadWriteOncePod (RWOP)

정답:

Question No : 3

Which prerequisite must be validated to ensure Nutanix Enterprise AI workloads can utilize GPU acceleration?

A.Controller VMs must be configured with GPU passthrough enabled by default.
B.Prism Central must be running version 2020.11 or later.
C.Each node must have dual 100GbE network interfaces.
D.Nodes must have supported GPUs and appropriate NVIDIA drivers installed.

정답:

Question No : 4

An administrator has successfully completed the Create an Endpoint wizard and returned to the Endpoint widget to note that the new endpoint displays a status of Failed.
The administrator does not see any obvious error message associated with the failed status and decides to engage Nutanix Support.
Which command output should be provided when raising the Nutanix Support case?

A.kubectl describe isvc endpoint_name -n nai-system
B.kubectl describe isvc endpoint_name -n nai-admin
C.kubectl exec isvc endpoint_name -n nai-admin
D.kubectl exec isvc endpoint_name -n nai-system

정답:

Question No : 5

An NAI administrator has successfully imported a model from Hugging Face and created an endpoint for the model. The endpoint is in the Active state. From within the Endpoint section in NAI, the endpoint has been tested with a Sample Request, the response is accurate, and the Status shows Succeeded.
The administrator has provided the endpoint URL and generated and provided API keys to the developers. However, the developers are having issues connecting to the endpoint. They keep getting 400 Bad Request errors when attempting to prompt the model.
What should the administrator do next to ensure the developers are able to successfully prompt the model?

A.Provide the developers the 'Sample Request' code from the endpoint by clicking 'View Sample Request'.
B.Check the Status of the endpoint again to make sure it was not accidentally Hibernated.
C.Re-import the model from Hugging Face, create a new endpoint and provide new URL and API keys to the developers.
D.Add additional resources to 'scale up' the number of instances of the endpoint.

정답:

Question No : 6

What is the correct endpoint PATH that is displayed in the NAI Dashboard?

A./api/v2/chat/completions
B./api/v1/chat/completions
C./api/v1/chat/endpoints
D./api/chat

정답:

Question No : 7

An administrator has installed and deployed NAI, but the UI is not launching.
Which command should the administrator use to ensure that all pods are successfully deployed?

A.kubectl get pods -n nai-system
B.kubectl get pods -n knative-serving
C.kubectl get pods -n kube-system
D.kubectl get pods -n ntnx-base

정답:

Question No : 8

An administrator observes that inference requests to Nutanix Enterprise AI are experiencing increased latency. After accessing the Infrastructure Usage and Health page, the administrator finds that the Service Health status is marked as Critical, and the GPU utilization on one node is consistently at 100%, while other nodes show moderate usage.
What is the most appropriate next step to resolve the performance issue?

A.Rebuild the endpoint to include multiple instances.
B.Restart the Kubernetes cluster to reset all node metrics and clear the Critical status.
C.Disable inference endpoints on the overloaded node.
D.Increase the memory allocation for the node with high GPU usage.

정답:

Question No : 9

Before installation, what kind of Kubernetes StorageClass must be provisioned for model files in an NFS shares for persistent volumes?

A.RWO
B.RWX
C.S3 RWX
D.S3 RWO

정답:

Question No : 10

A development team is building a new AI-powered customer service chatbot. They have successfully integrated an LLM from Hugging Face into their application via a Nutanix Enterprise AI (NAI) endpoint. The NAI administrator has confirmed the endpoint is in the 'Active' state, and internal sample requests run from the NAI Endpoint section return an 'accurate' response and 'Succeeded' status.
However, the developers report that the chatbot frequently provides vague or irrelevant answers to user queries, leading to low customer satisfaction scores. They describe the LLM's performance for this specific chatbot task as "suboptimal."
What should the administrator advise the developers to do next to improve the quality of the LLM's responses for the customer service chatbot task?

A.Instruct the administrator to check the Kubernetes pod logs for the endpoint to identify any CRASHLOOPBACKOFF or OOMKilled errors that might be affecting model inference.
B.Increase the number of instances for the NAI endpoint to enhance performance and reduce latency.
C.Reseed the initial dataset and reimport the existing LLM within Nutanix Enterprise AI without implementing a RAG platform.
D.Consult the Hugging Face documentation for the specific LLM to review its training data, intended use cases, and capabilities.

정답:

Question No : 11

What prerequisites must be skipped when adding GPU nodes to a managed Kubernetes service on an Azure AKS cluster using Azure CLI?

A.Install the GPU driver
B.Install the Worker driver
C.Install the Worker Connector
D.Install the GPU Connector

정답:

Question No : 12

Which task is an AI/ML User unable to perform in Nutanix Enterprise AI?

A.Create an API key to access an AI inference endpoint.
B.Delete an endpoint that they previously created.
C.Test an AI inference endpoint that they previously created.
D.View a LLM that was imported by a different user.

정답:

Question No : 13

What security requirement must a private container registry meet before pushing Nutanix AI
images?

A.Serve images over HTTPS with Basic-authentication only
B.Serve images over SSH with key-pair authentication required
C.Serve images over HTTP without trusted TLS/SSL certificates
D.Serve images over HTTPS with trusted TLS/SSL certificates

정답:

Question No : 14

An AI/ML admin is testing access to an endpoint using Open AI compatible clients, but is unable to successfully access the endpoint.
What could be the issue?

A.API key has been created by another AI/ML admin.
B.Endpoint has not been configured with the right LL
C.API key has been removed from the endpoint.
D.Endpoint as been created by another AI/ML admin.

정답:

Question No : 15

The administrator observes a "Pod Network Unreachable" health check failure in the NAI, specifically affecting several application pods on a newly added worker node. Existing pods on other nodes are functioning normally.
Upon initial investigation, kubectl get pods -o wide shows the affected pods stuck in ContainerCreating or Pending status, and kubectl describe pod <pod-name> events include messages, such as:
"FailedCreatePodSandBox: failed to create sandbox: rpc error: code = Unknown desc = failed to set up network for sandbox..." or "network is not ready"
What is the reason for this error and the appropriate next step that the administrator should take?

A.The Docker/containerd runtime on the worker node is corrupted; reinstall the container runtime.
B.The kube-proxy service is misconfigured; restart all kube-proxy pods on the cluster.
C.The underlying Nutanix network segmentation is incorrect; verify VLAN or IP pool configurations are correct in Prism Central.
D.The NKP CNI plugin on the affected worker node is not functioning correctly; inspect the logs of the CNI po

정답:

Nutanix NCP-AI 시험