Grafana Dashboards

DPF Grafana Dashboards

The DPF operator Helm chart packages Grafana dashboards as ConfigMaps that are deployed alongside the operator. The dashboard JSON definitions are located in the Helm chart under deploy/charts/dpf-operator/dashboards/.

Deployment

When grafanaDashboards.enabled is set to true in the Helm values, the chart creates two ConfigMaps:

• dpf-operator-grafana-dashboards: contains the fleet health, DPU detail, framework state, DPUCluster control plane performance, and DPUService performance dashboards. The ConfigMap carries the annotation grafana_folder: "DOCA Platform Framework" so the Grafana sidecar groups all of these dashboards in a dedicated DOCA Platform Framework folder.

• dpf-operator-grafana-debug-dashboards: contains the controller-runtime and framework performance dashboards. The annotation grafana_folder: "DOCA Platform Framework/debug" nests them in a debug subfolder under DOCA Platform Framework.

Both ConfigMaps are labeled with grafana_dashboard: "1", which allows the Grafana sidecar to automatically discover and load them. No manual import is required, once Grafana is running with the sidecar enabled, the dashboards appear automatically.

Dashboards

DPU Fleet Health

Provides a high-level overview of the entire DPU fleet including provisioning state, health conditions, DPUService readiness (desired, healthy, not-ready replicas), and cluster status.

DPU Detail

Provides detailed per-DPU information including resource utilization, networking, service status, and the list of pods scheduled on the selected DPU node (with DPUService attribution and readiness).

Framework State

Tracks the state of DPF custom resources across the cluster including DPUSets, DPUServices (with per-service node count), DPUDeployments, and provisioning objects.

Alerts

Summarizes the DPF Prometheus alerts at a glance: total active, firing, pending, critical, and warning counts, a filterable table of active alerts, a per-alert state timeline, and readiness-ratio snapshots. The alert counts require the Prometheus rule examples to be deployed; the readiness ratios additionally require the dpf-aggregates recording rules.

Unlike the dashboards above, this one is not part of the dpf-operator Helm chart, since it only makes sense together with the user-managed Prometheus rules. The deploy/helmfiles/monitoring.yaml helmfile deploys it automatically, together with those rules, via postsync hooks on the kube-prometheus-stack release. To deploy it into an existing monitoring stack manually, create a sidecar-discovered ConfigMap from the JSON:

kubectl -n dpf-operator-system create configmap dpf-monitoring-grafana-dashboards \
  --from-file=deploy/helmfiles/dashboards/ --dry-run=client -o yaml \
  | kubectl label --local -f - grafana_dashboard="1" --dry-run=client -o yaml \
  | kubectl annotate --local -f - grafana_folder="DOCA Platform Framework" --dry-run=client -o yaml \
  | kubectl apply --server-side -f -

or run the same helper the helmfile hook uses: deploy/helmfiles/apply-dashboards.sh [namespace].

DPUCluster Control Plane Performance

Surfaces the health and performance of the Kubernetes control plane running inside a DPUCluster: API server request rate and latency, inflight requests, active watch connections, controller-manager workqueue depth and latency, scheduler pending pods and attempt outcomes, and etcd request latency and storage size. Select the target DPUCluster via the cluster variable at the top of the dashboard.

DPUService Performance

Presents CPU and memory working-set usage of DPUService workloads running on DPU clusters: cluster-level capacity overview, per-namespace CPU and memory working set, and top pods by CPU and memory. The data comes from the container, pod, and node metrics that the DPU cluster OpenTelemetry Collector streams to the management cluster, so it requires metrics streaming to be enabled. Filter by DPU cluster, node, and namespace via the cluster, node, and namespace variables at the top of the dashboard.

Framework Performance

Monitors condition and phase transition performance of DPF custom resources such as DPUs, DPUServices, DPUClusters, and others.

Last updated: October 14, 2026