Grafana Dashboards¶

GitOps-managed observability dashboards deployed via Grafana Operator CRDs with Kustomize, providing real-time monitoring for AI workloads and infrastructure.

Overview¶

Grafana Dashboards is a dedicated repository for managing Grafana dashboards as code using Grafana Operator Custom Resource Definitions (GrafanaFolder, GrafanaDashboard). Dashboards are defined as Kubernetes CRDs with embedded JSON, deployed via Kustomize into the monitoring namespace, and automatically reconciled by the Grafana Operator running in the fleet-infra cluster.

This GitOps approach ensures dashboards are version-controlled, reviewable, and automatically synced to Grafana without manual UI configuration.

Key Features¶

Feature	Description
GitOps Managed	Dashboards defined as code, deployed via Flux CD sync
Grafana Operator CRDs	Native Kubernetes resources (GrafanaFolder, GrafanaDashboard) — not file provisioning
Fleet-Infra Integration	Targets the Grafana instance managed by kube-prometheus-stack
Folder Organization	Dashboards organized into GrafanaFolder resources by project
Auto-Refresh	Dashboards configured with live refresh intervals (30s default)
Kustomize Deployment	Composable resource management with Kustomize overlays

Technology Stack¶

Category	Technologies
Dashboard Engine	Grafana (via Grafana Operator)
CRD Types	GrafanaFolder, GrafanaDashboard
Deployment	Kustomize, Flux CD
Monitoring Backend	Prometheus (kube-prometheus-stack)
Platform	Kubernetes (fleet-infra cluster)
Namespace	`monitoring`

Quick Links¶

Dashboards - Detailed documentation for all dashboards and panels

Repository Structure¶

grafana-dashboards/
└── infra/
    ├── kustomization.yaml                  # Resource list: folder.yaml + dashboards
    ├── folder.yaml                         # GrafanaFolder CRD (agentic-ai folder)
    └── dashboards/
        └── agentic-ai/
            ├── dashboard.yaml              # GrafanaDashboard CRD with embedded JSON
            └── llm-metrics.json            # Dashboard JSON source

Dashboard Summary¶

Dashboard	Panels	Datasource	Folder	Tags
LLM Metrics Dashboard	14	Prometheus	agentic-ai	`llm`, `ai`, `performance`

Deployment Flow¶

flowchart LR
    subgraph "Developer"
        GIT[Git Push]
    end

    subgraph "GitOps"
        FLUX[Flux CD<br/>Source Sync]
        KUST[Kustomize<br/>Resource Assembly]
    end

    subgraph "Kubernetes (monitoring namespace)"
        GROP[Grafana Operator<br/>CRD Reconciliation]
        GRAF[Grafana Instance<br/>Dashboard Rendered]
    end

    GIT -->|Push to repo| FLUX
    FLUX -->|Sync manifests| KUST
    KUST -->|Apply CRDs| GROP
    GROP -->|Reconcile| GRAF

Integration with Fleet Infrastructure¶

The dashboards target the Grafana instance deployed by fleet-infra's kube-prometheus-stack. The Grafana Operator watches for CRDs matching the instance selector label dashboards: grafana.

flowchart TB
    subgraph "grafana-dashboards repo"
        FOLDER[GrafanaFolder CRD<br/>agentic-ai]
        DASH[GrafanaDashboard CRD<br/>LLM Metrics]
    end

    subgraph "fleet-infra cluster"
        subgraph "monitoring namespace"
            GROP[Grafana Operator]
            GRAF[Grafana<br/>kube-prometheus-stack]
            PROM[Prometheus<br/>Metrics Backend]
        end

        subgraph "agentic-ai namespace"
            APP[Agentic AI Backend<br/>Custom Metrics]
        end
    end

    FOLDER -->|Applied via Flux| GROP
    DASH -->|Applied via Flux| GROP
    GROP -->|Creates folder & dashboard| GRAF
    GRAF -->|Queries| PROM
    APP -->|Exposes metrics| PROM

CRD Instance Selector¶

All dashboard CRDs use the following instance selector to target the correct Grafana deployment:

instanceSelector:
  matchLabels:
    dashboards: grafana

This label matches the Grafana instance deployed by kube-prometheus-stack in fleet-infra.

Source Code¶

:octicons-mark-github-16: View on GitHub