Skip to main content

Dashboard Building — Assembling the Instrument Cluster

The mechanic has the dashboard frame. Now it's time to install the gauges. Building a Grafana dashboard is about choosing the right panel type, writing the right query, and arranging everything so the driver (engineer) sees the right information at a glance.

Creating Your First Dashboard​

  1. Click + → New Dashboard
  2. Click Add visualization
  3. Select Prometheus as the data source
  4. Enter a PromQL query
  5. Choose a visualization type
  6. Click Apply
Remember

Start with the four golden signals: Latency, Traffic, Errors, Saturation. One panel for each gives you a complete overview.

Panel Types — Choosing the Right Gauge​

Time Series — The Speedometer​

Shows values changing over time. This is the most common panel type.

Request rate over time
sum(rate(http_requests_total[5m])) by (method)

Best for: Anything that changes over time — CPU, memory, request rate, latency.

Stat — The Odometer​

Shows a single current value with optional color thresholds.

Current error count
sum(http_requests_total{status=~"5.."})

Best for: Key metrics you want to see at a glance — total errors, uptime percentage, active connections.

Gauge — The Fuel Gauge​

Shows a value within a min/max range with color bands.

CPU usage percentage
100 - (avg(rate(node_cpu_seconds_total{mode="idle"}[5m])) * 100)

Settings:

  • Min: 0, Max: 100
  • Thresholds: green < 70, yellow 70-90, red > 90

Best for: Percentages — CPU usage, memory usage, disk usage.

Table — The Diagnostic Report​

Shows structured data in rows and columns.

Top 10 endpoints by request count
topk(10, sum(rate(http_requests_total[5m])) by (path))

Best for: Comparing many items — top endpoints, slowest queries, largest pods.

Bar Gauge — The Comparison Chart​

Compares values across categories.

Requests per service
sum(rate(http_requests_total[5m])) by (service)

Best for: Comparing values across labels — requests per service, memory per pod, CPU per node.

Remember

Match the panel type to the question. "How is it changing?" → Time series. "What is it now?" → Stat. "How do they compare?" → Bar gauge.

Query Tips — Writing Better PromQL​

Use legends for readability​

Legend makes the graph readable
sum(rate(http_requests_total[5m])) by (method)
# Legend: {{method}}   → shows "GET", "POST", etc.

Multiple queries in one panel​

Compare two metrics
# Query A: Request rate
sum(rate(http_requests_total[5m]))

# Query B: Error rate
sum(rate(http_requests_total{status=~"5.."}[5m]))

Math operations between queries​

Error percentage (query B / query A)
# Query A: total requests
# Query B: error requests
# Then use: B / A * 100
Remember

Legends turn cryptic metric names into human-readable labels. Always add a legend to your queries.

Thresholds — Color-Coding Health​

Thresholds change panel colors based on values:

Threshold configuration
thresholds:
  steps:
    - color: green
      value: null       # 0 to first threshold
    - color: yellow
      value: 70         # 70% to 90%
    - color: red
      value: 90         # 90% and above
ColorMeaning
GreenHealthy
YellowWarning — investigate soon
RedCritical — act now
Remember

Thresholds are the visual equivalent of alert rules. They make problems visible at a glance.

Dashboard Organization — Best Practices​

1. Top row: Overview panels​

Put the most important metrics at the top. Use Stat panels for:

  • Total requests
  • Error rate
  • Average latency
  • Uptime

2. Group by signal type​

  • Row 1: Overview (stat panels)
  • Row 2: Traffic (request rate graphs)
  • Row 3: Latency (response time graphs)
  • Row 4: Errors (error rate graphs)
  • Row 5: Saturation (CPU, memory, disk)

3. Use variables for multi-service dashboards​

Variables
templating:
  list:
    - name: service
      type: query
      query: label_values(http_requests_total, service)
    - name: instance
      type: query
      query: label_values(http_requests_total{service="$service"}, instance)
Remember

A well-organized dashboard answers questions in order: "Is everything healthy?" → "Which service has a problem?" → "What's the root cause?"

Export and Import — Sharing Dashboards​

Export a dashboard as JSON
# In Grafana UI: Dashboard → Settings → JSON Model → Copy
Import a dashboard
# In Grafana UI: + → Import → Paste JSON or enter dashboard ID

Grafana has a community dashboard library at grafana.com/grafana/dashboards. Popular IDs:

  • Node Exporter Full: 1860
  • Docker: 893
  • Kubernetes: 315
Remember

Don't build from scratch if a community dashboard exists. Import it and customize.