Grafana — Metric Visualisation and Dashboards

Metrics in Prometheus are useful, but raw numbers do not tell anyone whether the system is healthy. Grafana visualises metrics in real time — graphs, heatmaps, alerting. An operational dashboard you actually want to look at.

Installation and Datasources¶

Grafana runs as a Docker container. Datasource: Prometheus, Elasticsearch, InfluxDB, PostgreSQL — connect any data source. A single dashboard can combine data from multiple sources.

Dashboard Design¶

Rules for effective dashboards:

• Overview first: Top row — RED metrics (Rate, Errors, Duration)
• Drill-down: Click on a graph → detail per service
• Variables: Dropdown for selecting service, environment, time range
• Alerting: Visual indicator — green/yellow/red

Template Variables¶

# Prometheus query for a variable
label_values(up, job)

# Usage in a panel
rate(http_requests_total{job="$service"}[5m])

One dashboard, a dropdown to select the service. No need to create a dashboard for each service separately.

Alerting¶

Grafana alerting or Prometheus Alertmanager. Rules: error rate > 5%, p95 response time > 1s, disk usage > 80%. Notifications: Slack, email, PagerDuty.

A Dashboard is a Window into the System¶

A good dashboard saves hours of debugging. Invest in dashboard design the same way you invest in code. Grafana community dashboards are a great starting point.