Telemetry and Logging

NVIDIA-managed services automatically deploy a ServiceMonitor resource to allow metrics collection by a Kubernetes Prometheus stack. To disable this, set monitoring.enabled=false in the Helm values file for each deployment.

By default, the ServiceMonitor is deployed in the same namespace as the service. To specify a different namespace, set the monitoring.prometheusNamespace value.

For alternative monitoring setups, Omniverse services export Prometheus and OpenTelemetry-compatible metrics. Prometheus metrics are available at the /metrics endpoint, with default rate, error, and duration metrics exposed.