Observability#
Platform Monitoring#
Platform-level monitoring for GPU, SR-IOV networking, Precision Time Protocol, and NMOS Registry services is provided by Holoscan for Media. Refer to Platform Monitoring in the Holoscan for Media user guide.
Kubernetes Status#
Check deployment rollout:
kubectl rollout status deployment/<deployment-name> -n <namespace> --timeout=300s
Check pod status and events:
kubectl get pods -n <namespace>
kubectl describe pod <pod-name> -n <namespace>
Stream VSR NIM logs:
kubectl logs -n <namespace> deployment/<deployment-name> --follow
On Red Hat OpenShift, replace kubectl with oc.
Operator Status#
For operator deployments, inspect the custom resource:
kubectl get nvidiavsrnimmediafunction -n <namespace>
kubectl describe nvidiavsrnimmediafunction <cr-name> -n <namespace>
The following are useful status fields:
status.conditionsstatus.inputs.video_input.controlstatus.outputs.video_output.controlstatus.managed_resources
The operator updates status from the generated deployment and, in NMOS mode, from the NMOS HTTP API on the VSR NIM pod.
NMOS#
For NMOS deployments, verify device registration and connection state through the NMOS Registry and Controller. NMOS Registry monitoring is covered in the Holoscan for Media platform monitoring guide.
Metrics#
The VSR NIM ST 2110 media function image in this release does not currently expose a Prometheus /metrics endpoint.
The Helm chart metrics.enabled setting and the operator CR parameters.metricsEnabled setting add Prometheus scrape annotations only. Leave media-function metrics disabled until the deployed image exposes a compatible endpoint.
The operator controller metrics are separate from media-function metrics. The operator chart can expose controller-runtime metrics when operator metrics are enabled.