splunk-observability-nvidia-gpu-integration

You can collect NVIDIA GPU metrics in Splunk Observability Cloud by rendering a custom Prometheus receiver and metrics pipeline that ingests DCGM Exporter telemetry across Kubernetes clusters, supporting both GPU Operator and standalone deployments.

To avoid Splunk OTel chart autodetect collisions when adding NVIDIA GPU telemetry, use a uniquely named receiver_creator (dcgm-cisco by default) and a dedicated metrics pipeline to prevent chart integration conflicts.

Yes, you can track per-workload GPU metrics in Splunk Observability by optionally patching DCGM pod labels to expose workload-level metadata, enabling detailed telemetry correlation for Kubernetes deployments.

Yes, the NVIDIA GPU integration supports both GPU Operator deployments and legacy DCGM exporters by using dual-label pod discovery to find and ingest telemetry across different Kubernetes cluster configurations.

Prebuilt dashboards and detectors for NVIDIA GPU telemetry in Splunk Observability are generated alongside the rendered overlay, providing immediate visibility into GPU health and performance metrics without manual configuration.

You can validate a DCGM Exporter receiver overlay against your Kubernetes cluster using the provided setup workflow script, which outputs a ready-to-apply overlay and patch artifacts for your Splunk Observability deployment.