Grafana — Overview#

Identity#

  • Module path: github.com/grafana/grafana
  • Go version: 1.25.8
  • License: AGPL-3.0-only (with Apache-2.0 exceptions documented in LICENSING.md)
  • Repository: https://github.com/grafana/grafana

Purpose#

Grafana is an open-source monitoring and observability platform that lets teams query, visualize, alert on, and understand metrics, logs, and traces from virtually any data source. It targets DevOps engineers, SREs, and data teams who need a unified, dashboard-driven view across heterogeneous infrastructure. The platform supports mixed data sources per graph, templated dynamic dashboards, and rule-based alerting with notifications to Slack, PagerDuty, OpsGenie, and others.

Significance#

Grafana is one of the most widely adopted observability tools in the cloud-native ecosystem, with millions of installations and a thriving plugin ecosystem. It serves as the de facto visualization layer in the LGTM stack (Loki, Grafana, Tempo, Mimir) championed by Grafana Labs. Its influence extends through deeply integrated open-source projects (Prometheus, InfluxDB, Elasticsearch), enterprise offerings, and a large contributor community. The project’s architectural decisions around plugin isolation (gRPC), CUE-based schema generation, and Wire-based DI at massive scale make it a reference implementation for Go monorepo patterns.

Key metrics#

  • Go files: ~5,867 (actual count; work item listed size tier L, actual is XL territory)
  • Top-level directories: apps/, conf/, contribute/, cue.mod/, devenv/, docs/, e2e/, e2e-playwright/, emails/, grafana-mixin/, hack/, kinds/, kindsv2/, packages/, packaging/, pkg/, public/, scripts/, tools/
  • Direct dependencies: ~258 (from go.mod; annotated with owning team per entry)
  • First commit / age: Shallow clone — git history not fully available; Grafana was open-sourced in January 2014, making it ~12 years old

Notable characteristics#

  • True monorepo at scale: Combines a Go backend (pkg/), a TypeScript/React frontend (public/), standalone App SDK apps (apps/), Yarn workspaces for frontend packages (packages/), and CUE schemas (kinds/) — all versioned together with coordinated build tooling.
  • Wire dependency injection at enterprise scale: The backend wires hundreds of services via Google Wire, providing compile-time dependency graph validation and eliminating entire classes of runtime init bugs. Regenerated with make gen-go after any service init change.
  • Plugin system via gRPC/protobuf: Plugins (data sources, panels, apps) communicate with the backend over gRPC, providing language-agnostic extensibility and strong isolation. Plugin workspaces are independent Yarn packages with separate build steps.
  • CUE-driven schema generation: Dashboard and panel type definitions live in kinds/ as CUE schemas and generate both Go structs and TypeScript types, ensuring frontend/backend type parity without manual synchronization.
  • Dependency ownership via comments: Every direct Go dependency in go.mod is annotated with its owning Grafana team (e.g., // @grafana/alerting-backend), providing a live ownership map inside the module manifest — an unusually disciplined practice for a project of this size.
Backward Prometheus — Testing Grafana — Structure Forward