Dashboard Design

Dashboard Design

Design judgment for Grafana dashboards — what to build and how it should look. For authoring mechanics (MCP workflow, JSON templates, ConfigMap), see the grafana-dashboards skill. For alert and ServiceMonitor authoring, see monitoring-authoring.

Design Workflow

Define purpose → Layout hierarchy → Select panel types → Apply color/thresholds
     → Push via MCP → Screenshot review → Iterate → Hand off to grafana-dashboards

Step 1: Define Dashboard Purpose

One dashboard = one question it answers. Examples:

• "Is the platform healthy right now?" → Platform Home
• "Why is latency elevated?" → Service-level signals
• "How full is storage?" → Storage capacity

If you can't state the question in one sentence, split into multiple dashboards.

Step 2: Layout Hierarchy

Structure rows top-to-bottom by audience urgency (Z-pattern reading flow):

Grid widths (24-column grid):

• 4 stats per row: w:6 at x: 0, 6, 12, 18
• 3 panels per row: w:8 at x: 0, 8, 16
• 6 compact stats: w:4 at x: 0, 4, 8, 12, 16, 20
• Full-width: w:24

Step 3: Panel Type Selection

Stat panels dominate this codebase (54 stat : 42 timeseries : 18 gauge). Default to stat for KPIs, timeseries for anything with meaningful time shape.

Step 4: Color & Threshold Design

Two threshold modes — pick based on which direction is bad:

More-is-worse (errors, utilization, latency):

"steps": [
  { "color": "green", "value": null },
  { "color": "yellow", "value": 70 },
  { "color": "red", "value": 90 }
]

Less-is-worse (availability %, SLO compliance):

"steps": [
  { "color": "red", "value": null },
  { "color": "yellow", "value": 50 },
  { "color": "green", "value": 90 }
]

Binary (on/off, present/absent):

"mappings": [{ "type": "value", "options": {
  "0": { "text": "DOWN", "color": "red" },
  "1": { "text": "UP", "color": "green" }
}}]

Domain thresholds from this codebase:

Step 5: Visual Review Loop

Use MCP get_panel_image for quick panel review during authoring:

MCP: get_panel_image  dashboardUid: "uid"  panelId: 1  width: 800  height: 400

Use Playwright for full-dashboard layout review (spacing, row flow, color consistency):

1. kubectl port-forward svc/grafana 3000:80 -n monitoring
2. Navigate to http://localhost:3000/d/<uid>
3. browser_take_screenshot → inspect layout, color, readability
4. Fix issues, push via MCP update_dashboard, re-screenshot

Check for: crowded rows, inconsistent threshold colors, missing panel titles, axes without units, buried golden signals.

Anti-Patterns

Keywords

visual design, dashboard layout, panel hierarchy, information architecture, panel type selection, color semantics, thresholds, Z-pattern, golden signals, SLO, screenshot review, iterative design, Playwright, visual feedback, information density, stat panel, timeseries, gauge, table, bar gauge, value mapping, color conventions, grid layout