Grafana Dashboards

Dashboards are provisioned automatically from ConfigMaps. Create a ConfigMap with your dashboard JSON, apply the correct label, and Grafana picks it up.

Creating a tenant dashboard ConfigMap

apiVersion: v1
kind: ConfigMap
metadata:
  name: observability-dashboards-tenant-workloads
  namespace: observability
  labels:
    kupe.cloud/grafana-dashboards: "true"
    kupe.cloud/grafana-folder: "Workloads"
data:
  my-app-overview.json: |-
    {
      "title": "My App Overview",
      "uid": "my-app-overview",
      ...
    }

Requirements

The label kupe.cloud/grafana-dashboards: "true" is required for operator-driven tenant dashboard sync.
Use kupe.cloud/grafana-folder: "<Folder Name>" to control target folder (for example, Overview or Workloads).
Place ConfigMaps in the Grafana namespace (observability by default).
Each key in data must end in .json.
The uid field in the dashboard JSON must be unique across all dashboards.
Do not use grafana_dashboard or grafana_folder for tenant dashboards. Those are sidecar labels for Main Org dashboards and can route dashboards to the wrong place.

Folder organization

Tenant dashboards are organized by ConfigMap label, not filename prefix.

apiVersion: v1
kind: ConfigMap
metadata:
  name: observability-dashboards-tenant-overview
  namespace: observability
  labels:
    kupe.cloud/grafana-dashboards: "true"
    kupe.cloud/grafana-folder: "Overview"
data:
  service-health.json: |-
    { ... }

Mapping examples:

kupe.cloud/grafana-folder: "Overview" -> Overview
kupe.cloud/grafana-folder: "Workloads" -> Workloads
No folder label -> General

Folders are created automatically in each tenant org if they do not exist.

Building dashboard JSON

The easiest way to create dashboard JSON:

Build your dashboard in the Grafana UI.
Click the Share icon and select Export.
Toggle Export for sharing externally to replace data source references with variables.
Copy the JSON and paste it into your ConfigMap.

Minimum viable dashboard

{
  "title": "My Dashboard",
  "uid": "unique-id",
  "version": 1,
  "schemaVersion": 39,
  "refresh": "30s",
  "time": { "from": "now-1h", "to": "now" },
  "tags": ["my-app"],
  "templating": {
    "list": [
      {
        "name": "datasource",
        "type": "datasource",
        "query": "prometheus",
        "refresh": 1
      },
      {
        "name": "namespace",
        "type": "query",
        "datasource": { "type": "prometheus", "uid": "${datasource}" },
        "query": "label_values(kube_pod_info, namespace)",
        "refresh": 2,
        "includeAll": true
      }
    ]
  },
  "panels": [
    {
      "title": "Request Rate",
      "type": "timeseries",
      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 0 },
      "datasource": { "type": "prometheus", "uid": "${datasource}" },
      "targets": [
        {
          "expr": "sum(rate(http_requests_total{namespace=\"$namespace\"}[5m])) by (pod)",
          "legendFormat": "{{ pod }}"
        }
      ]
    }
  ]
}

Data source reference

Use a datasource variable for tenant dashboards:

"datasource": { "type": "prometheus", "uid": "${datasource}" }

With this templating block:

{
  "name": "datasource",
  "type": "datasource",
  "query": "prometheus"
}

Dashboard design tips

Use variables for namespace, workload, and cluster to make dashboards reusable.
Keep time range and resolution appropriate to the investigation window.
Pair rate and error metrics on one panel for faster correlation.
Set meaningful thresholds with color coding (green/yellow/red).
Use stat panels for key numbers and timeseries for trends.

Deploying via Helm

If you manage dashboards via Helm, render them into tenant dashboard ConfigMaps with the Kupe labels:

{{- if .Values.dashboard.enabled }}
apiVersion: v1
kind: ConfigMap
metadata:
  name: {{ .Values.dashboard.configMapName | default "observability-dashboards-tenant-workloads" }}
  namespace: {{ .Values.dashboard.configMapNamespace | default "observability" }}
  labels:
    kupe.cloud/grafana-dashboards: "true"
    kupe.cloud/grafana-folder: {{ .Values.dashboard.folder | default "Workloads" | quote }}
data:
{{ (.Files.Glob "dashboards/*.json").AsConfig | indent 2 }}
{{- end }}

Place dashboard JSON files in a dashboards/ directory alongside your chart templates.

Default dashboards

Kupe provides baseline tenant dashboards out of the box:

Overview / Cluster — cluster health and capacity overview.
Workloads / Pod Resources — per-pod CPU, memory, network, and filesystem usage.

Platform admins also have additional dashboards in Main Org (overview, workloads, system, infrastructure, billing).

Review cadence

Review dashboard usefulness after incidents.
Remove low-signal panels.
Add panels for recurring failure patterns.