Knowing that Grafana is a query-and-display layer explains where data, performance and permissions really live.
Grafana is a web application for visualizing and alerting on data that lives elsewhere. It connects to data sources such as Prometheus, Loki (logs), Tempo (traces), Elasticsearch, PostgreSQL, MySQL, InfluxDB, CloudWatch or Azure Monitor, sends them queries and draws the results.
The main concepts:
- Data source: a configured connection with a URL and credentials.
- Dashboard: a page of panels arranged in a grid, with a shared time range and variables (drop-downs) at the top.
- Panel: one or more queries plus a visualization such as a time series graph, stat, gauge, table, heatmap or logs view.
- Explore: an ad-hoc query view for investigations, without building a dashboard.
- Alerting: rules that evaluate queries and notify through contact points.
Grafana itself stores only its own configuration (dashboards, users, alert rules) in a database, SQLite by default. Your metrics stay in Prometheus, so a slow panel is usually a slow query in the data source.
Grafana is open source (AGPLv3). Grafana Cloud offers it as a hosted service together with Prometheus-compatible storage (Mimir), Loki and Tempo, and Grafana Enterprise adds features like fine-grained access control and reporting.
Remember
- Grafana queries data sources and visualizes results; it doesn't store metrics
- Dashboards contain panels; panels contain queries and a visualization
- Explore is for ad-hoc investigation