Skip to content

Deploying Nutanix Enterprise AI (NAI) NVD Reference Application

Version 2.3.0

This version of the NAI deployment is based on the Nutanix Enterprise AI (NAI) v2.3.0 release.

stateDiagram-v2
    direction LR

    state DeployNAI {
        [*] --> DeployNAIAdmin
        DeployNAIAdmin -->  InstallSSLCert
        InstallSSLCert --> DownloadModel
        DownloadModel --> CreateNAI
        CreateNAI --> [*]
    }

    [*] --> PreRequisites
    PreRequisites --> DeployNAI 
    DeployNAI --> TestNAI : next section
    TestNAI --> [*]

Prepare for NAI Deployment

Changes in NAI v2.4.0

  • Istio Ingress gateway is replaced with Envoy Gateway
  • Knative is removed from NAI
  • Kserve has been upgraded to 0.15.0

Enable NKP Applications

Enable these NKP Applications from NKP GUI.

Note

In this lab, we will be using the Management Cluster Workspace to deploy our Nutanix Enterprise AI (NAI)

However, in a customer environment, it is recommended to use a separate workload NKP cluster.

Info

The helm charts for these applications are stores in NKP's Chart Museum The container images are stored in internal Harbor registry. These images got uploaded to Harbor at the time of install NKE in this section.

  1. In the NKP GUI, Go to Clusters
  2. Click on Management Cluster Workspace
  3. Go to Applications

  4. Search and enable the following applications: follow this order to install dependencies for NAI application

    • Kube-prometheus-stack: version 70.4.2 or later (pre-installed on NKP cluster)

  5. Login to VSC on the jumphost VM, append the following environment variables to the $HOME\airgap-nai\.env file and save it

    export ENVIRONMENT=nkp
export NAI_USER=_your_desired_nai_username
export NAI_TEMP_PASS=_your_desired_nai_password # At least 8 characters

    export ENVIRONMENT=nkp
export NAI_USER=admin
export NAI_TEMP_PASS=_XXXXXXXXX # At least 8 characters

  6. IN VSC,go to Terminal and run the following commands to source the environment variables

    source $HOME/airgap-nai/.env

  7. Enable Envoy Gateway v1.5.0 using the following command

    helm install envoy-gateway \
  oci://${REGISTRY_HOST}/gateway-helm \
  --version v1.5.0 \
  --set image.repository=$REGISTRY_HOST/envoyproxy/gateway \
  --set image.tag=v1.5.0 \
  -n envoy-gateway-system \
  --create-namespace

    Pulled: harbor.10.x.x.111.nip.io/nkp/gateway-helm:v1.5.0
Digest: sha256:2435a9cfcf22043b5ea2cdfe1e5783ec81f1dc527bff3c46c80c3ecc3ed66915
NAME: envoy-gateway
LAST DEPLOYED: Fri Aug 29 06:08:29 2025
NAMESPACE: envoy-gateway-system
STATUS: deployed
REVISION: 1
TEST SUITE: None

  8. Check if Envoy Gateway resources are ready

    kubectl wait --timeout=5m -n envoy-gateway-system deployment/envoy-gateway --for=condition=Available

    deployment.apps/envoy-gateway condition met

  9. Run the Kserve CRD installation

    helm install kserve-crd \
  oci://${REGISTRY_HOST}/kserve-crd \
  --version v0.15.0 \
  -n kserve \
  --create-namespace

  10. Run the Kserve installation

    helm install kserve \
  oci://${REGISTRY_HOST}/kserve \
  --version v0.15.0 \
  -n kserve \
  --set controller.image.repository=${REGISTRY_HOST}/kserve-controller \
  --set controller.image.tag=v0.15.0 \
  --set kserve.controller.deploymentMode=RawDeployment \
  --set kserve.controller.gateway.disableIngressCreation=true

    Pulled: harbor.apj-cxrules.win/nkp/kserve:v0.15.0
Digest: sha256:dae9d3e35c96d318bf3f5eb15e303ed6c268129f988f0b0b4ad698dc414d3f40
NAME: kserve
LAST DEPLOYED: Tue Sep  2 01:56:48 2025
NAMESPACE: kserve
STATUS: deployed
REVISION: 1
TEST SUITE: None
    Check helm deployment status

    Check the status of the nai helm deployments using the following command:

    helm list -n envoy-gateway-system
helm list -n kserve

Deploy NAI

  1. Source the environment variables (if not done so already)

    source $HOME/airgap-nai/.env

  2. In VSCode Explorer pane, browse to $HOME/airgap-nai folder

  3. Run the following command to create a helm values file:

    cat << EOF > ${ENVIRONMENT}-values.yaml
naiIepOperator:
  iepOperatorImage:
    image: ${REGISTRY_HOST}/nai-iep-operator
    tag: ${NAI_API_VERSION}
  modelProcessorImage:
    image: ${REGISTRY_HOST}/nai-model-processor
    tag: ${NAI_API_VERSION}
naiInferenceUi:
  naiUiImage:
    image: ${REGISTRY_HOST}/nai-inference-ui
    tag: ${NAI_API_VERSION}
naiApi:
  naiApiImage:
    image: ${REGISTRY_HOST}/nai-api
    tag: ${NAI_API_VERSION}
  supportedKserveRuntimeImage: ${REGISTRY_HOST}/nai-kserve-huggingfaceserver
  supportedKserveCPURuntimeImageTag: ${KSERVE_VERSION}
  supportedKserveGPURuntimeImageTag: ${KSERVE_VERSION}-gpu
  supportedKserveCustomModelServerRuntimeImage: ${REGISTRY_HOST}/nai-kserve-custom-model-server
  supportedKserveCustomModelServerRuntimeImageTag: ${NAI_API_VERSION}
  # Details of super admin (first user in the nai system)
  superAdmin:
    username: ${NAI_USER}
    password: ${NAI_TEMP_PASS} # At least 8 characters
    # email: admin@nutanix.com
    # firstName: admin
#${REGISTRY_HOST}/nai-kserve-huggingfaceserver:kserve-version
  supportedTGIImage: ${REGISTRY_HOST}/nai-tgi
  supportedTGIImageTag: ${NAI_TGI_RUNTIME_VERSION}
naiDatabase:
  naiDbImage:
    image: ${REGISTRY_HOST}/nai-postgres:${NAI_POSTGRESQL_VERSION}
naiMonitoring:
  prometheus:
    image: 
      registry: ${REGISTRY_HOST}
      repository: prometheus/prometheus
      tag: ${NAI_PROMETHEUS_VERSION}
EOF

    naiIepOperator:
  iepOperatorImage:
    image: harbor.10.x.x.111.nip.io/nkp/nai-iep-operator
    tag: v2.4.0
  modelProcessorImage:
    image: harbor.10.x.x.111.nip.io/nkp/nai-model-processor
    tag: v2.4.0
naiInferenceUi:
  naiUiImage:
    image: harbor.10.x.x.111.nip.io/nkp/nai-inference-ui
    tag: v2.4.0
naiApi:
  naiApiImage:
    image: harbor.10.x.x.111.nip.io/nkp/nai-api
    tag: v2.4.0
  supportedKserveRuntimeImage: harbor.10.x.x.111.nip.io/nkp/nai-kserve-huggingfaceserver
  supportedKserveCPURuntimeImageTag: "v0.15.2"
  supportedKserveGPURuntimeImageTag: "v0.15.2-gpu"
  supportedKserveCustomModelServerRuntimeImage: harbor.10.x.x.111.nip.io/nkp/nai-kserve-custom-model-server
  supportedKserveCustomModelServerRuntimeImageTag: "v2.4.0"
# harbor.10.x.x.111.nip.io/nkp/nai-kserve-huggingfaceserver:kserve-version
  supportedTGIImage: harbor.10.x.x.111.nip.io/nkp/nai-tgi
  supportedTGIImageTag: "3.3.4-b2485c9"
  # Details of super admin (first user in the nai system)
  superAdmin:
    username: admin         
    password: _XXXXXXXXXX # At least 8 characters
    # email: admin@nutanix.com
    # firstName: admin
naiDatabase:
  naiDbImage:
    image: harbor.10.x.x.111.nip.io/nkp/nai-postgres:16.1-alpine
naiMonitoring:
  prometheus:
    image: 
      registry: harbor.10.x.x.111.nip.io/nkp
      repository: prometheus/prometheus
      tag: v2.54.0

  4. In VSCode, Under $HOME/airgap-nai folder, click on New File and create a file with the following name:

    nai-deploy.sh

    with the following content:

    #!/usr/bin/env bash

set -ex
set -o pipefail

helm upgrade --install nai-core --repo oci://${REGISTRY_HOST}/nai-core \
--version=${NAI_CORE_VERSION} -n nai-system --create-namespace \
--insecure-skip-tls-verify \
-f ${ENVIRONMENT}-values.yaml --wait

  5. Run the following command to deploy NAI

    $HOME/airgap-nai/nai-deploy.sh

    $HOME/airgap-nai/nai-deploy.sh 

Release "nai-core" does not exist. Installing it now.
Pulled: harbor.10.x.x.111.nip.i/nkp/nai-core:2.4.0
Digest: sha256:283b8373ca76088d89fbf91482ef8530f6608f30085f84863e17edae77efe673
NAME: nai-core
LAST DEPLOYED: Tue Sep  2 01:31:15 2025
NAMESPACE: nai-system
STATUS: deployed
REVISION: 1
TEST SUITE: None

  6. Verify that the NAI Core Pods are running and healthy

    kubens nai-system
kubectl get po,deploy

    $ kubens nai-system
✔ Active namespace is "nai-system"

$ kubectl get po,deploy

NAME                                           READY   STATUS      RESTARTS   AGE
pod/nai-api-db-migrate-tt6rn-cp5fc             0/1     Completed   1          7m
pod/nai-api-fbc4f956d-h5vk2                    1/1     Running     0          7m
pod/nai-db-0                                   1/1     Running     0          7m
pod/nai-iep-model-controller-f84596945-ck5tk   1/1     Running     0          7m
pod/nai-ui-d5b546bfc-82x6l                     1/1     Running     0          7m
pod/prometheus-nai-0                           2/2     Running     0          7m

NAME                                       READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/nai-api                    1/1     1            1           7m
deployment.apps/nai-iep-model-controller   1/1     1            1           7m
deployment.apps/nai-ui                     1/1     1            1           7m
Uninstall NAI v2.3.0 Dependencies

If you are upgrading NAI from v2.3.0 to v2.4.0, uninstall the following:

If Helm was used:

Uninstall Istio
helm uninstall istio-ingressgateway -n istio-system --wait --ignore-not-found
helm uninstall istiod -n istio-system --wait --ignore-not-found
helm uninstall istio-base -n istio-system --wait --ignore-not-found
Uninstall Knative
kubectl delete --ignore-not-found=true KnativeServing knative-serving -n knative-serving
helm uninstall knative-operator -n knative-serving --wait --ignore-not-found
kubectl wait --for=delete pod --all -n knative-serving --timeout=300s

If NKP Application were used for installation:

Go to NKP Cluster Dashboard > Application > Search and Uninstall the following:

  1. Istio
  2. Knative

Install SSL Certificate and Gateway Elements

In this section we will install SSL Certificate to access the NAI UI. This is required as the endpoint will only work with a ssl endpoint with a valid certificate.

NAI UI is accessible using the Ingress Gateway.

The following steps show how cert-manager can be used to generate a self signed certificate using the default selfsigned-issuer present in the cluster.

If you are using Public Certificate Authority (CA) for NAI SSL Certificate

If an organization generates certificates using a different mechanism then obtain the certificate + key and create a kubernetes secret manually using the following command:

kubectl -n istio-system create secret tls nai-cert --cert=path/to/nai.crt --key=path/to/nai.key

Skip the steps in this section to create a self-signed certificate resource.

  1. Get the NAI UI ingress gateway host using the following command:

    NAI_UI_ENDPOINT=$(kubectl get svc -n envoy-gateway-system -l "gateway.envoyproxy.io/owning-gateway-name=nai-ingress-gateway,gateway.envoyproxy.io/owning-gateway-namespace=nai-system" -o jsonpath='{.items[0].status.loadBalancer.ingress[0].ip}' | grep -v '^$' || kubectl get svc -n envoy-gateway-system -l "gateway.envoyproxy.io/owning-gateway-name=nai-ingress-gateway,gateway.envoyproxy.io/owning-gateway-namespace=nai-system" -o jsonpath='{.items[0].status.loadBalancer.ingress[0].hostname}')

  2. Get the value of NAI_UI_ENDPOINT environment variable

    echo $NAI_UI_ENDPOINT

    10.x.x.216

  3. We will use the command output e.g: 10.x.x.216 as the IP address for NAI as reserved in this section

  4. Construct the FQDN of NAI UI using nip.io and we will use this FQDN as the certificate's Common Name (CN).

    nai.${NAI_UI_ENDPOINT}.nip.io

    nai.10.x.x.216.nip.io

  5. Create the ingress resource certificate using the following command:

    cat << EOF | k apply -f -
apiVersion: cert-manager.io/v1
kind: Certificate
metadata:
  name: nai-cert
  namespace: nai-system
spec:
  issuerRef:
    name: selfsigned-issuer
    kind: ClusterIssuer
  secretName: nai-cert
  commonName: nai.${NAI_UI_ENDPOINT}.nip.io
  dnsNames:
  - nai.${NAI_UI_ENDPOINT}.nip.io
  ipAddresses:
  - ${NAI_UI_ENDPOINT}
EOF

  6. Patch the Envoy gateway with the nai-cert certificate details

    kubectl patch gateway nai-ingress-gateway -n nai-system --type='json' -p='[{"op": "replace", "path": "/spec/listeners/1/tls/certificateRefs/0/name", "value": "nai-cert"}]'

  7. Create EnvoyProxy

    k apply -f -<<EOF
apiVersion: gateway.envoyproxy.io/v1alpha1
kind: EnvoyProxy
metadata:
  name: envoy-service-config
  namespace: nai-system
spec:
  provider:
    type: Kubernetes
    kubernetes:
      envoyService:
        type: LoadBalancer
EOF

  8. Patch the nai-ingress-gateway resource with the new EnvoyProxy details

    kubectl patch gateway nai-ingress-gateway -n nai-system --type=merge \
-p '{
    "spec": {
        "infrastructure": {
            "parametersRef": {
                "group": "gateway.envoyproxy.io",
                "kind": "EnvoyProxy",
                "name": "envoy-service-config"
            }
        }
    }
}'

Accessing the UI

  1. In a browser, open the following URL to connect to the NAI UI

    https://nai.10.x.x.216.nip.io

  2. Use the ${NAI_USER} and ${NAI_TEMP_PASS} values set in ${ENVIRONMENT}-values.yaml files during helm installation of NAI v.2.4.0

  3. Change the password for the admin user

  4. Login using admin user and password.

Download Model

We will download and user llama3 8B model which we sized for in the previous section.

  1. In the NAI GUI, go to Models
  2. Click on Import Model from Hugging Face
  3. Choose the meta-llama/Meta-Llama-3.1-8B-Instruct model

  4. Input your Hugging Face token that was created in the previous section and click Import

  5. Provide the Model Instance Name as Meta-Llama-3.1-8B-Instruct and click Import

  6. Go to VSC Terminal to monitor the download

    Get jobs in nai-admin namespace
    kubens nai-admin

kubectl get jobs
    Validate creation of pods and PVC
    kubectl get po,pvc
    Verify download of model using pod logs
    kubectl logs -f _pod_associated_with_job

    Get jobs in nai-admin namespace
    kubens nai-admin

✔ Active namespace is "nai-admin"

kubectl get jobs

NAME                                       COMPLETIONS   DURATION   AGE
nai-c0d6ca61-1629-43d2-b57a-9f-model-job   0/1           4m56s      4m56
    Validate creation of pods and PVC
    kubectl get po,pvc

NAME                                             READY   STATUS    RESTARTS   AGE
nai-c0d6ca61-1629-43d2-b57a-9f-model-job-9nmff   1/1     Running   0          4m49s

NAME                                       STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS      VOLUMEATTRIBUTESCLASS   AGE
nai-c0d6ca61-1629-43d2-b57a-9f-pvc-claim   Bound    pvc-a63d27a4-2541-4293-b680-514b8b890fe0   28Gi       RWX            nai-nfs-storage   <unset>                 2d
    Verify download of model using pod logs
    kubectl logs -f nai-c0d6ca61-1629-43d2-b57a-9f-model-job-9nmff 

/venv/lib/python3.9/site-packages/huggingface_hub/file_download.py:983: UserWarning: Not enough free disk space to download the file. The expected file size is: 0.05 MB. The target location /data/model-files only has 0.00 MB free disk space.
warnings.warn(
tokenizer_config.json: 100%|██████████| 51.0k/51.0k [00:00<00:00, 3.26MB/s]
tokenizer.json: 100%|██████████| 9.09M/9.09M [00:00<00:00, 35.0MB/s]<00:30, 150MB/s]
model-00004-of-00004.safetensors: 100%|██████████| 1.17G/1.17G [00:12<00:00, 94.1MB/s]
model-00001-of-00004.safetensors: 100%|██████████| 4.98G/4.98G [04:23<00:00, 18.9MB/s]
model-00003-of-00004.safetensors: 100%|██████████| 4.92G/4.92G [04:33<00:00, 18.0MB/s]
model-00002-of-00004.safetensors: 100%|██████████| 5.00G/5.00G [04:47<00:00, 17.4MB/s]
Fetching 16 files: 100%|██████████| 16/16 [05:42<00:00, 21.43s/it]:33<00:52, 9.33MB/s]
## Successfully downloaded model_files|██████████| 5.00G/5.00G [04:47<00:00, 110MB/s] 

Deleting directory : /data/hf_cache

  7. Optional - verify the events in the namespace for the pvc creation 

    k get events | awk '{print $1, $3}'

    $ k get events | awk '{print $1, $3}'

3m43s Scheduled
3m43s SuccessfulAttachVolume
3m36s Pulling
3m29s Pulled
3m29s Created
3m29s Started
3m43s SuccessfulCreate
90s   Completed
3m53s Provisioning
3m53s ExternalProvisioning
3m45s ProvisioningSucceeded
3m53s PvcCreateSuccessful
3m48s PvcNotBound
3m43s ModelProcessorJobActive
90s   ModelProcessorJobComplete

The model is downloaded to the Nutanix Files pvc volume.

After a successful model import, you will see it in Active status in the NAI UI under Models menu

Create and Test Inference Endpoint

In this section we will create an inference endpoint using the downloaded model.

  1. Navigate to Inference Endpoints menu and click on Create Endpoint button

  2. Fill the following details:

    • Endpoint Name: llama-8b
    • Model Instance Name: Meta-LLaMA-8B-Instruct
    • Use GPUs for running the models : Checked
    • No of GPUs (per instance):
    • GPU Card: NVIDIA-L40S (or other available GPU)
    • No of Instances: 1
    • API Keys: Create a new API key or use an existing one

  3. Click on Create

  4. Monitor the nai-admin namespace to check if the services are coming up

    kubens nai-admin
kubectl get po,deploy

    kubens nai-admin
get po,deploy
NAME                                                     READY   STATUS        RESTARTS   AGE
pod/llama8b-predictor-00001-deployment-9ffd786db-6wkzt   2/2     Running       0          71m

NAME                                                 READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/llama8b-predictor-00001-deployment   1/1     1            0           3d17h

  5. Check the events in the nai-admin namespace for resource usage to make sure all 

    kubectl get events -n nai-admin --sort-by='.lastTimestamp' | awk '{print $1, $3, $5}'

    $ kubectl get events -n nai-admin --sort-by='.lastTimestamp' | awk '{print $1, $3, $5}'

110s FinalizerUpdate Updated
110s FinalizerUpdate Updated
110s RevisionReady Revision
110s ConfigurationReady Configuration
110s LatestReadyUpdate LatestReadyRevisionName
110s Created Created
110s Created Created
110s Created Created
110s InferenceServiceReady InferenceService
110s Created Created

  6. Once the services are running, check the status of the inference service

    kubectl get isvc

    kubectl get isvc

NAME      URL                                          READY   PREV   LATEST   PREVROLLEDOUTREVISION   LATESTREADYREVISION       AGE
llama8b   http://llama8b.nai-admin.svc.cluster.local   True           100                              llama8b-predictor-00001   3d17h

Troubleshooting Endpoint ISVC

TGI Imange and Self-signed Certificates

Only follow this procedure if this isvc is not starting up.

KNative Serving Image Tag Checking

From testing, we have identified that KServe module is making sure that there are no container image tag discrepencies, by pulling image using SHA digest. This is done to avoid pulling images that are updated without updating the tag.

We have avoided this behavior by patching the config-deployment config map in the knative-serving namespace to skip image tag checking. Check this Prepare for NAI Deployment sectionfor more details.

kubectl patch configmap  config-deployment -n knative-serving --type merge -p '{"data":{"registries-skipping-tag-resolving":"${REGISTRY_HOST}"}'

If this procedure was not followed, then the isvc will not start up.

  1. If the isvc is not coming up, then explore the events in nai-admin namespace.

    kubens nai-admin
kubectl get isvc
kubectl get events  --sort-by='.lastTimestamp'

    $ kubectl get isvc

NAME      URL                                          READY   PREV   LATEST   PREVROLLEDOUTREVISION   LATESTREADYREVISION       AGE
llama8b   http://llama8b.nai-admin.svc.cluster.local   False

$ kubectl get events --sort-by='.lastTimestamp'

Warning   InternalError         revision/llama8b-predictor-00001   Unable to fetch image "harbor.10.x.x.111.nip.io/nkp/nutanix/nai-tgi:2.3.1-825f39d": failed to resolve image to digest: 
Get "https://harbor.10.x.x.111.nip.io/v2/": tls: failed to verify certificate: x509: certificate signed by unknown authority

    The temporary workaround is to use the TGI images SHA signature from the container registry.

    This site will be updated with resolutions for the above issues in the future.

  2. Note the above TGI image SHA digest from the container registry.

    docker pull ${REGISTRY_HOST}/nutanix/nai-tgi:${NAI_TGI_RUNTIME_VERSION}

    docker pull harbor.10.x.x.111.nip.io/nkp/nutanix/nai-tgi:2.3.1-825f39d

2.3.1-825f39d: Pulling from nkp/nutanix/nai-tgi
Digest: sha256:2df9fab2cf86ab54c2e42959f23e6cfc5f2822a014d7105369aa6ddd0de33006
Status: Image is up to date for harbor.10.x.x.111.nip.io/nkp/nutanix/nai-tgi:2.3.1-825f39d
harbor.10.x.x.111.nip.io/nkp/nutanix/nai-tgi:2.3.1-825f39d

  3. The SHA digest will look like the following:

    TGI image SHA digest will be different for different environments
    sha256:2df9fab2cf86ab54c2e42959f23e6cfc5f2822a014d7105369aa6ddd0de33006

  4. Create a copy of the isvc manifest

    kubectl get isvc llama8b -n nai-admin -o yaml > llama8b.yaml

  5. Edit the isvc

    kubectl edit isvc llama8b -n nai-admin

  6. Search and replace the image tag with the SHA digest from the TGI image.

    <snip>

env:
- name: STORAGE_URI
  value: pvc://nai-c34d8d58-d6f8-4cb4-94e4-28-pvc-claim/model-files
  image: harbor.10.x.x.111.nip.io/nkp/nutanix/nai-tgi:2.3.1-825f39d

<snip>
    7. After replacing the image's SHA digest, the image value should look as follows: 

    <snip>

env:
- name: STORAGE_URI
  value: pvc://nai-c34d8d58-d6f8-4cb4-94e4-28-pvc-claim/model-files
  image: harbor.10.x.x.111.nip.io/nkp/nutanix/nai-tgi@sha256:2df9fab2cf86ab54c2e42959f23e6cfc5f2822a014d7105369aa6ddd0de33006

<snip>

  7. Save the isvc configuration by writing the changes to the file and exiting the vi editor using :wq! key combination.

  8. Verify that the isvc is running

    kubens nai-admin
kubectl get isvc

    $ kubectl get isvc

NAME      URL                                          READY   PREV   LATEST   PREVROLLEDOUTREVISION   LATESTREADYREVISION       AGE
llama8b   http://llama8b.nai-admin.svc.cluster.local   True           100                              llama8b-predictor-00001   3d17h

This should resolve the issue the issue with the TGI image.

Report Other Issues

If you are facing any other issues, please report them here in the NAI LLM GitHub Repo Issues page.