Grafana¶
Summary
Grafana is the de facto open-source platform for querying, visualizing, and alerting on observability data from almost any source. Grafana Labs builds it in Go (backend) and TypeScript/React (frontend) under AGPL-3.0-only. It is the "G" in the LGTM stack (Loki, Grafana, Tempo, Mimir, plus Pyroscope for profiles). The current line is Grafana 13 (April 2026). It made Git Sync (bidirectional dashboards-as-code) and dynamic dashboards (schema v2) generally available, moved resources onto a Kubernetes-style API with unified storage, and brought the Grafana Assistant AI agent to self-managed installs. Telemetry collection is done by Grafana Alloy; Grafana Agent reached end of life on 2025-11-01.
Key Facts¶
| Attribute | Detail |
|---|---|
| Latest Version | 13.2.2 (2026-09-15); 13.2.0 released 2026-08-18 |
| Supported lines | 13.2, 13.1, 13.0, and 12.4 (extended, last 12.x minor); see Reference |
| Release cadence | Major yearly (April–May), minor every other month, monthly patches; each minor is supported for 9 months (last minor of a major gets 15) |
| License | AGPL-3.0-only since Grafana 8.0 (April 2021; Apache-2.0 before). SDK packages, Alloy, and the Helm charts are Apache-2.0 |
| Governance | Single-vendor open source, owned by Grafana Labs (not a CNCF project) |
| Company | Grafana Labs, founded 2014 (as Raintank), HQ New York; $400M+ ARR and 7,000+ customers (2025-09); reportedly raising at ~$9B valuation (2026-02) |
| Creator | Torkel Ödegaard (first release January 2014, forked from the Kibana 3 frontend) |
| Repository | github.com/grafana/grafana: 73.1k stars and 13.7k forks as of 2026-04 |
| Languages | TypeScript and Go |
| Collector | Grafana Alloy 1.20.0 (2026-09-25), Apache-2.0 |
| Install path | grafana/grafana image, apt.grafana.com packages, or the grafana-community/grafana Helm chart; managed via Grafana Cloud, Amazon Managed Grafana, or Azure Managed Grafana |
| Pricing | OSS free; Grafana Cloud Free ($0), Pro ($19/month + usage), Enterprise (from ~$25k/year) |
Architecture at a Glance¶
Grafana sits between users and data sources: it stores only configuration and forwards every query to the backend that owns the data. The diagram shows the main parts; more detail is in Explanation.
flowchart LR
Alloy["Grafana Alloy<br/>(collector)"] --> Backends
subgraph Backends["Data sources"]
Mimir["Mimir / Prometheus"]
Loki["Loki"]
Tempo["Tempo"]
Pyro["Pyroscope"]
Other["SQL, CloudWatch,<br/>Elasticsearch, 100+ more"]
end
subgraph GF["Grafana server"]
Q["Query service<br/>+ plugins (gRPC)"]
AL["Alerting engine"]
US["Unified storage<br/>/apis resources"]
end
Git["Git repo<br/>(Git Sync)"] <--> US
US --> DB[("MySQL / PostgreSQL<br/>(SQLite for dev)")]
Q --> Backends
AL --> Q
Users["Browser / gcx / Terraform"] --> GF
What Changed in 2025–2026¶
| When | Change |
|---|---|
| 2025-03 | Grafana OnCall OSS enters maintenance mode (archived 2026-03-24; replacement is Grafana Cloud IRM) |
| 2025-05 | Grafana 12.0: AngularJS removed; Git Sync, dashboard schema v2, and dynamic dashboards introduced as experimental; Drilldown apps GA; SQL expressions; alert Recovering state |
| 2025-07 | Grafana 12.1: API key endpoints removed (service accounts only) |
| 2025-10 | Grafana Assistant GA in Grafana Cloud; Assistant Investigations preview |
| 2025-11 | Grafana Agent end of life (2025-11-01); critical CVE-2025-41115 (Enterprise SCIM) patched |
| 2026-01/03 | Grafana, Loki (OSS), and Tempo Helm charts move to grafana-community/helm-charts |
| 2026-04 | Grafana 13.0 (GrafanaCON): Git Sync and dynamic dashboards GA, Grafana Advisor GA, Assistant on-prem, unified storage migration, /api deprecated in favour of /apis, grafana-cli/grafana-server removed. 13.0.0 was withdrawn; use 13.0.1+ |
| 2026-06 | Grafana 13.1: Git Sync verified commits and folderless sync, quick filters, Assistant pre-installed in Enterprise; grafanactl archived in favour of gcx |
| 2026-07 | "AI Week": Assistant Investigations, Automations, Workspace, Cloud MCP server, gcx, Agent Observability GA |
| 2026-08 | Grafana 13.2: saved queries GA, Git Sync for GitHub Enterprise plus GitLab/Bitbucket webhooks, scripted dashboards deprecated |
Evaluation¶
Why It Stands Out¶
Vendor-neutral and open-standards-first (Prometheus, OpenTelemetry), with the broadest data-source catalogue of any open tool (100+ data sources through plugins). It can put metrics, logs, traces, and profiles into one UI with click-through correlation. No other open-source tool combines this breadth with an active managed offering.
When It Fits¶
- You need dashboards and alerting over heterogeneous sources (Prometheus plus SQL plus cloud monitoring).
- You want to avoid backend lock-in: visualization is portable, and data stays in your stores.
- You run or plan the LGTM stack, or Grafana Cloud.
- You want dashboards and alerts as code reviewed in Git (Git Sync,
gcx, Terraform).
It fits less well when you want a single-vendor, zero-ops APM product with automatic instrumentation and no data-source setup (Datadog, New Relic), or a log-search-first workflow (Kibana/OpenSearch Dashboards).
Pros and Cons¶
| Pros | Cons |
|---|---|
| Unmatched data-source flexibility | Self-hosting the full LGTM stack is operationally heavy |
| Strong visualization engine (dynamic dashboards, dozens of panel types) | AGPL-3.0 can create compliance friction for some organizations |
| Large community and plugin catalogue | Unified alerting has a steep learning curve |
| Open standards (Prometheus, OTel native) | Default SQLite backend is dev-only; HA needs MySQL/PostgreSQL |
| Generous Grafana Cloud free tier | Fine-grained RBAC, SAML, reporting, auditing are Enterprise/Cloud-only |
| Cross-signal correlation (metrics, logs, traces, profiles) | Dashboard sprawl is a real governance problem at scale |
| First-class GitOps since 13.0 (Git Sync) | Fast release pace with a breaking major every year (Angular in 12, CLI and renderer in 13) |
Common Use Cases¶
- Infrastructure and Kubernetes monitoring
- Application performance monitoring with Tempo and OpenTelemetry
- Log exploration with Loki
- Business and SQL dashboards (PostgreSQL, MySQL, BigQuery, Snowflake)
- IoT and industrial telemetry
- Security dashboards (Loki or Elasticsearch based)
- AI/LLM and agent observability (Grafana Cloud AI Observability, Agent Observability)
Licensing and Commercial Use¶
- Core Grafana: AGPL-3.0-only since April 2021. Running unmodified Grafana, even commercially, carries no copyleft obligations; if you modify it and offer it to users over a network, you must publish the modifications under AGPL.
- Apache-2.0 parts:
@grafana/data,@grafana/ui,@grafana/runtime, schemas and kinds, plugin SDKs, Alloy, Helm charts, Terraform provider. - Grafana Enterprise: commercial license on top of the same codebase (RBAC, SAML, reporting, auditing, enterprise plugins, Assistant pre-installed).
- Grafana Cloud: Free, Pro ($19/month platform fee + usage, e.g. ~$6.50 per 1k metric series), Enterprise (from ~$25k/year). Details and caveats are in Reference.
- Managed alternatives: Amazon Managed Grafana (per-user pricing), Azure Managed Grafana.
Ecosystem and Compatibility¶
- Native backends: Prometheus, Mimir, Loki, Tempo, Pyroscope.
- First-party data sources: Elasticsearch, OpenSearch, InfluxDB, MySQL, PostgreSQL, MS SQL, CloudWatch, Azure Monitor, Google Cloud Monitoring, BigQuery, Jaeger, Zipkin, Graphite, and more; Enterprise adds Snowflake, Splunk, Datadog, ServiceNow, and others.
- Collection: Grafana Alloy (OTel Collector distribution; replaces Grafana Agent) and Beyla/OBI for eBPF auto-instrumentation.
- As code: Git Sync,
gcx, Foundation SDK, file provisioning, Terraform provider, Ansible collection, Grafana Operator, Crossplane. - APIs: legacy REST
/api(deprecated in 13.0) and Kubernetes-style/apis; backend plugins over gRPC. - Runs on: Linux (Debian/Ubuntu, RHEL/Fedora, SUSE), macOS, Windows, Docker, Kubernetes.
- Database: SQLite (default, dev only), MySQL 8.0+, PostgreSQL 12+.
- Sizing: official small tier is 2 cores and 2–4 GB RAM (fewer than 25 concurrent users); see Reference.
Alternatives¶
| Alternative | Positioning vs Grafana |
|---|---|
| Datadog | All-in-one SaaS with faster time to value, much higher cost at scale |
| Kibana / OpenSearch Dashboards | Stronger for log-search-centric ELK workflows |
| New Relic | SaaS APM with a generous free tier |
| SigNoz | Open-source, OpenTelemetry-native all-in-one (ClickHouse) |
| OpenObserve | Open-source all-in-one with built-in UI |
| Coroot | eBPF-based, zero-instrumentation observability with built-in UI |
| Perses | CNCF sandbox dashboards-as-code project for Prometheus-style data |
| Apache Superset | BI-focused, less real-time |
| Splunk Observability | Enterprise suite, expensive |
Migration and Lock-in¶
- Low lock-in on the visualization layer: dashboards are JSON or v2 resources; data sources are external and swappable.
- Moderate lock-in with the full LGTM stack: LogQL and TraceQL are Grafana-specific, though storage uses open formats (Prometheus TSDB blocks, Parquet) in your own object storage, and ingestion is OTLP/Prometheus remote-write.
- Moving in: Grafana documents migrations from Datadog, New Relic, and Splunk to Grafana Cloud, plus a Cloud Migration Assistant (GA in 12.0) from self-managed to Cloud.
- Moving out: Git Sync and
gcx resources pullexport resources as files; Terraform state documents everything else.
Community Health and Support¶
- Very active repository: patch releases for four supported lines every few weeks, frequent
+securityreleases, and contributors across Grafana Labs and the community. - Community: community.grafana.com, Grafana Slack, GrafanaCON (annual, where majors ship) and ObservabilityCON.
- Commercial support: Grafana Cloud Pro (8x5 email) and Enterprise SLAs.
- Adoption signal: Grafana Labs reports 7,000+ paying customers (2025-09) and a Leader position in the Gartner Magic Quadrant for Observability Platforms (2024).
Topic Map¶
- How-to Guides: install, production configuration, HA, authentication, hardening, Git Sync and
gcx, provisioning, Terraform, Alloy pipelines, API recipes, upgrades, troubleshooting. - Reference: versions and support windows, editions and pricing, licensing, requirements, ports,
grafana.inikeys, CLI, APIs, roles, alert states, Git Sync limits, metrics, Helm charts, deprecations, advisories, hardening checklist. - Explanation: architecture, request lifecycle, plugins, app platform and unified storage, schema v2, Git Sync design, alerting, LGTM integration, Alloy, AI features, security model, history.
Related Topics¶
- LGTM vs Victoria Stack
- Observability Stacks Comparison
- Observability comparisons index
- LGTM Stack: Loki, Tempo, Mimir, and Pyroscope internals, deployment, and benchmarks
- OpenTelemetry: the instrumentation standard Alloy implements
- OpenTelemetry Collector: the upstream collector Alloy is a distribution of
- VictoriaMetrics: alternative metrics backend commonly used with Grafana
- Observability 2.0: the wide-events critique of the three-pillars model
- eBPF Developer Tutorial: the instrumentation layer behind Beyla/OBI-style auto-instrumentation
- MySQL: supported Grafana backend database and data source
Sources¶
| URL | Kind | Used for |
|---|---|---|
| https://github.com/grafana/grafana | Repository | Code, license |
| https://github.com/grafana/grafana/blob/main/CHANGELOG.md | Changelog | Versions, dates, breaking changes |
| https://github.com/grafana/grafana/blob/main/LICENSING.md | License map | AGPL vs Apache paths |
| https://grafana.com/docs/grafana/latest/ | Docs | General reference |
| https://grafana.com/docs/grafana/latest/whatsnew/whats-new-in-v13-0/ | Release notes | Grafana 13 features |
| https://grafana.com/docs/grafana/latest/upgrade-guide/upgrade-v13.0/ | Upgrade guide | 13.0 breaking changes, withdrawn 13.0.0 |
| https://grafana.com/docs/grafana/latest/upgrade-guide/when-to-upgrade/ | Docs | Release cadence and support policy |
| https://grafana.com/docs/grafana/latest/setup-grafana/installation/ | Docs | Requirements and databases |
| https://grafana.com/docs/grafana/latest/as-code/observability-as-code/git-sync/ | Docs | Git Sync |
| https://grafana.com/whats-new/2026-04-20-git-sync-for-grafana-dashboards-and-folders-now-generally-available/ | What's new | Git Sync GA |
| https://grafana.com/whats-new/2026-04-08-dynamic-dashboards-is-now-generally-available/ | What's new | Dynamic dashboards GA |
| https://grafana.com/docs/grafana/latest/alerting/ | Docs | Alerting |
| https://grafana.com/docs/grafana/latest/administration/provisioning/ | Docs | Provisioning |
| https://grafana.com/developers/plugin-tools/ | Docs | Plugin development |
| https://grafana.com/docs/alloy/latest/ | Docs | Alloy |
| https://github.com/grafana/alloy/blob/main/CHANGELOG.md | Changelog | Alloy versions |
| https://github.com/grafana/agent | Repository | Agent EOL notice |
| https://github.com/grafana/gcx | Repository | gcx CLI |
| https://github.com/grafana-community/helm-charts | Repository | Helm charts (new home) |
| https://github.com/grafana/terraform-provider-grafana | Repository | Terraform provider |
| https://grafana.com/pricing/ | Pricing | Grafana Cloud plans |
| https://monitoringcost.com/grafana-cloud-pricing | Third-party | 2026 per-unit prices (cross-check) |
| https://grafana.com/press/2025/10/08/grafana-labs-revolutionizes-ai-powered-observability-with-ga-of-grafana-assistant-and-introduces-assistant-investigations/ | Press | Assistant GA |
| https://grafana.com/press/2026/07/27/grafana-labs-ships-six-tools-that-power-agentic-operations-from-planning-to-production/ | Press | AI Week releases |
| https://grafana.com/press/2025/09/30/grafana-labs-surpasses-400m-arr-and-7000-customers-gains-new-investors-to-accelerate-global-expansion/ | Press | ARR and customers |
| https://siliconangle.com/2026/02/13/grafana-labs-reportedly-raising-funding-9b-valuation/ | News | Reported valuation |
| https://grafana.com/blog/2021/04/20/grafana-loki-tempo-relicensing-to-agplv3/ | Blog | AGPL relicensing |
| https://grafana.com/blog/grafana-enterprise-security-update-critical-severity-security-fix-for-cve-2025-41115/ | Advisory | CVE-2025-41115 |
| https://github.com/grafana/oncall | Repository | OnCall OSS archive notice |
| https://community.grafana.com/ | Forum | Community |
| https://play.grafana.org/ | Demo | Live demo instance |
Questions¶
Open Questions¶
- What exactly does Grafana Assistant send to Grafana Cloud when used from a self-managed (on-prem) instance, and which LLM vendors back it? (Grafana Labs had not disclosed the model vendor as of 2026-04.) See Explanation.
- When will the legacy
/apiendpoints be removed? 13.0 only deprecated them; a future major will announce removal. See Reference. - Will Git Sync limits (10 connections, ~1,000 resources per connection) rise as promised toward ~100 connections per stack? See Reference.
- Is Sift still a separate product, or fully folded into Assistant Investigations?
- Current Grafana Cloud per-unit prices should be re-checked directly on grafana.com/pricing (not fetchable during the 2026-09 review).
Answered Questions¶
- What license is Grafana under? AGPL-3.0-only since April 2021 (Grafana 8.0); Apache-2.0 before. See Reference.
- Can Grafana scale horizontally? Yes. Run several replicas against a shared MySQL or PostgreSQL database behind a load balancer. Sessions live in the database, so Redis is optional (remote cache, alerting HA). Configure alerting HA so notifications are deduplicated. See How-to Guides.
- What is Grafana Alloy? Grafana Labs' OpenTelemetry Collector distribution with Prometheus pipelines; the successor to Grafana Agent (EOL 2025-11-01). See Explanation.
- How does Loki differ from Elasticsearch? Loki indexes labels only, not full text, which makes storage far cheaper but requires label-first queries. See LGTM Stack.
- What are Grafana's AI features? Grafana Assistant (GA 2025-10 in Cloud, on-prem with Grafana 13), Assistant Investigations (multi-signal root-cause agents, GA 2026-07), plus Sift (earlier automated diagnostics). See Explanation.
- What does Grafana Cloud cost compared with self-hosted LGTM? Grafana Cloud Pro bills about $6.50 per 1,000 active series and about $0.45/GB for logs, traces, and profiles above the free allowances (10k series, 50 GB each), plus $19/month. Rough mid-size estimates: self-hosted ~$700–2,000/month infrastructure (plus operations staff), Grafana Cloud Pro ~$4,400–7,800+/month at list price for 500k–1M billable series and ~100 GB/day of logs (recomputed 2026-09-27; the earlier ~$1,000–3,300 range was below list price). Earlier notes stated "$8 per active series", which confused the unit (it is per 1,000). See Explanation.
- How does Mimir query sharding compare with Thanos? Mimir shards a query by series hash into N partial queries (
-query-frontend.query-sharding-total-shards, default 16) executed in parallel, while Thanos mainly splits by time range. Mimir's approach is faster for high-cardinality queries but uses more querier and store-gateway resources. Details belong to the LGTM Stack topic. - How do I migrate from ELK to Loki and Grafana? Dual-ship with Alloy, design low-cardinality labels, rebuild dashboards and rewrite queries in LogQL, keep Elasticsearch read-only until its retention expires. See How-to Guides.
- How production-ready is Pyroscope integration? Pyroscope is a multi-tenant microservices system (Pyroscope 2.0, April 2026, introduced a new v2 storage architecture), supports push SDKs (Go, Java, Python, Ruby, .NET, Rust, Node.js) and eBPF collection via Alloy, and links to traces through span profiles; Profiles Drilldown is part of Grafana. Considered production-viable. See LGTM Stack.
- Grafana Enterprise vs Grafana Cloud Enterprise? Same Enterprise feature set (RBAC, SAML/team sync, auditing, reporting, enterprise plugins, 24x7 support); Enterprise is self-managed, Cloud Enterprise is hosted with the backends included. See Reference.
- How does Scenes change dashboard and plugin development?
@grafana/scenesis the scene-graph framework that now renders all dashboards (schema v2 and v1) and lets app plugins build dashboard-like pages with query runners, transformations, variables, layouts, and URL sync. See Explanation. - Best Loki label strategy for 1,000+ Kubernetes pods? Keep about 10 or fewer low-cardinality labels (cluster, namespace, app, container), never
podor trace IDs as labels, and move high-cardinality fields to structured metadata (Loki 3.x). See How-to Guides. - TraceQL metrics vs the span-metrics generator? TraceQL metrics (introduced in Tempo 2.4, GA in Tempo 3.0 per the LGTM topic) compute metrics from traces at query time, which is flexible for exploration. The metrics-generator pre-computes RED metrics into a Prometheus-compatible store, which is faster for dashboards and alerts. See LGTM Stack.