← Back to glossary

Grafana & Dashboards

observability Beginner

TL;DR

Grafana is the de facto open-source dashboard platform — connecting to Prometheus, Loki, Elasticsearch, and 50+ data sources to visualise metrics, logs, and traces in a unified UI.

Explanation

Grafana: visualisation layer for any observability data. Panel types: time series, stat, gauge, histogram, table, heatmap, logs. Variables: dynamic dashboard filtering (environment, service, instance). Alerting: Grafana-managed alerts with multiple notification channels. Data sources: Prometheus, Loki, Elasticsearch, Tempo (traces), InfluxDB, CloudWatch, Datadog. Grafana Cloud: managed Grafana + Prometheus + Loki + Tempo. Best practices: one dashboard per service (golden signals), one overview dashboard per team, use variables for filtering. Export/import dashboards as JSON (share via grafana.com).

Common Misconception

✗ Grafana stores data — it only visualises. Grafana reads from data sources (Prometheus, Loki) but stores no metrics or logs itself.

Why It Matters

Grafana dashboards provide the operational UI that makes metrics and logs actionable — without visualisation, raw numbers in Prometheus are hard to interpret.

Common Mistakes

Too many panels per dashboard — aim for 10-15, not 50.
No variables for filtering — can't scope to one service or environment.
Dashboard sprawl — hundreds of dashboards, none maintained.
Using Grafana as the alerting source of truth — prefer Prometheus alerting rules.

Code Examples

✗ Vulnerable

# 50-panel dashboard with no variables:
# One fixed dashboard for all environments, all services

✓ Fixed

# Service dashboard with variables:
variables:
  - name: service
    query: label_values(up, job)
  - name: env
    query: label_values(up, env)

# Four golden signals panels:
# 1. Request rate
# 2. Error rate
# 3. P99 Latency
# 4. Saturation (CPU/Memory)

# Import from grafana.com/dashboards — e.g. Laravel dashboard #12345

References

↗ https://grafana.com/docs/grafana/latest/