Grafana

Grafana is open-source software for querying, visualising, alerting on and exploring metrics, logs and traces, wherever they are stored. It does not store most of that data itself. Instead it connects to data sources and presents them together.

How it works

A data source is a connection to a backend such as Prometheus for metrics, Loki for logs, or Tempo for traces. A panel runs a query against a data source and draws the result as a graph, table, gauge or similar. A dashboard arranges panels on one page, and variables let a single dashboard be reused across clusters, namespaces or services.

Loki and Tempo are separate Grafana Labs projects with their own documentation, which is why a Grafana installation can sit in front of a stack built from several independent systems. Alerting in Grafana lets you define alert rules against supported data sources and route notifications to contact points. Teams that already run Alertmanager with Prometheus often keep rules there and use Grafana mainly for dashboards.

A practical example

An engineer investigating slow requests opens a dashboard showing request latency from Prometheus. A spike on one service is visible, so they switch to the Explore view, query the logs for the same time range, and look for errors. Because both views share the time range, the correlation is quick.

Common mistakes

• Building dashboards by hand in the browser without exporting them. Store dashboards as code, such as JSON or provisioning files, in version control so they can be reviewed and restored.
• Packing every metric onto one dashboard. Start from the questions responders ask and build a small number of focused views.
• Alerting on every graph. Alerts should reflect user impact, ideally tied to SLIs and SLOs.
• Giving all users editor rights, and exposing data source credentials, instead of using roles and least privilege.

When to use it and when not to

Grafana is a good fit when you need one interface over several data sources, or when you want to avoid being tied to a single monitoring vendor. It is a poor fit if you expect it to collect and store data on its own: you still need collectors and backends. For logging concepts see log aggregation, and for managed delivery see our monitoring service.

See also

• Prometheus
• Alertmanager
• Log Aggregation
• Grafana technology page