Skip to navigation

End-to-End Verification

View as MarkdownOpen in Claude

Use this checklist after installing the nvidia-bnr-h4m-sample chart.

1. Verify Release Status

helm status bnr-h4m-sample

2. Verify Pods

kubectl get pods -o wide

Expected workloads:

  • Sender pod.
  • BNR NIM service pod.
  • Receiver pod.

On Red Hat OpenShift, replace kubectl with oc.

3. Verify Model Download

Check the BNR NIM pod logs for model download and Triton startup:

kubectl logs deploy/<nim-app-name>

If model download fails, verify that the generic model pull secret exists and contains the NGC_API_KEY key:

kubectl get secret ngc-api-key

4. Verify ST 2110 or NMOS Connectivity

For static ST 2110 mode, verify that the sender output IP address and port matches the BNR input and that the BNR output IP address and port matches the receiver input.

For NMOS mode, verify registration and connections in the NMOS Connection Manager UI:

  • BNR Audio Sender output connected to BNR-NIM input.
  • BNR-NIM output connected to BNR Audio Receiver input.

5. Preview Receiver Output

Find the node internal IP:

kubectl get nodes -o wide

Play the SRT preview stream:

ffplay "srt://<node-internal-ip>:32090"

6. Check for Common Issues

SymptomCheck
Image pull failureConfirm that the docker-registry image pull secret has access to nvcr.io.
Model download failureConfirm that ngc-api-key exists and contains key NGC_API_KEY.
Pod stuck pendingConfirm GPU availability, SR-IOV resources, scheduler, and node selector.
No media receivedConfirm Multus network, multicast IP/port chain, and NMOS connections.
Audio sample-rate mismatchConfirm that model.type or parameters.modelType matches the stream sample rate.