{"canonical_url":"https://engineer.company/glossary/grafana/","content_markdown":"# Grafana\n\n## What it is\n\nGrafana turns numbers kept in other systems into dashboards. It sends a query to\na data source such as Prometheus, Loki, PostgreSQL or MySQL, draws what comes\nback as a time-series graph, a table, a gauge or a heatmap, and raises an alert\nwhen a value crosses a line someone has set. Its own documentation puts it as\nsoftware to query, visualise, alert on and explore metrics, logs and traces\n\"wherever they are stored\", and the last four words carry the design.\n\nThe telemetry on the screen stays where it was collected. Each panel fetches its\nslice from the data source when the dashboard loads or refreshes, through one of\nmore than 150 data source plugins, and Grafana keeps only its own\nconfiguration: dashboards, folders, users, permissions and alert rules. Those\nlive in a small database that is SQLite by default, embedded in the binary, with\nMySQL or PostgreSQL recommended for production. A dashboard is a JSON document:\npanels arranged in rows, each panel a query plus a choice of how to draw it, and\nvariables along the top that let one dashboard switch between hosts,\nenvironments or services from a drop-down.\n\nGrafana Labs develops it, alongside Loki for logs, Tempo for traces and Mimir\nfor long-term Prometheus storage, a set people often call the Grafana stack. On\n20 April 2021 Grafana, Loki and Tempo moved from the Apache 2.0 licence to\nAGPLv3. Grafana comes as the open-source edition, as Grafana Enterprise under a\ncommercial licence, and as Grafana Cloud, which Grafana Labs hosts and runs.\n\nIt is usually weighed against Kibana, which plays the same part in front of\nElasticsearch, against the expression browser built into Prometheus, and\nagainst hosted observability suites that sell storage and screens as one\nproduct. Business-intelligence tools such as Metabase or Power BI can read the\nsame databases, but they are built for reports a person studies, while Grafana\nis built for screens that refresh every few seconds.\n\n## Why it matters\n\nInstalling Grafana is the easy part. It runs on a small machine and can be\npointed at a database a team already has. The decisions that shape what it is\nworth come afterwards: who each screen is for, how dashboards are kept once\nthere are dozens of them, and where it runs.\n\nBecause it only displays, adopting it moves no data. A team can keep Prometheus,\nPostgreSQL and its log store where they are, and could swap the display layer\nlater without migrating a single series, so the lock-in and the bill both sit\nwith the data sources. Suites that sell storage and display together make the\nopposite trade: one bill and one login, and data that is harder to take\nelsewhere.\n\nThe audience shapes a dashboard more than the tool does. An engineer on call\nwants a page that answers \"is it broken, and where\" at a glance, with links into\ndetail. A room of people with different jobs needs fewer and larger panels,\nplain titles and a threshold line that says when to worry. The same Grafana\nserves both through different dashboards, and a page built for both at once\ntends to serve neither.\n\nHow dashboards are kept is the second decision. Grafana can load them from JSON\nfiles on disk and watch the folder for changes, which its documentation calls\nprovisioning. In version control a dashboard gets review, history and a way\nback, and a new environment gets the same screens without anyone clicking\nthrough them. The cost is that editing in the browser stops being how changes\nare made, and the whole team has to accept that.\n\nLicence and hosting come last. Under AGPLv3, whoever modifies Grafana and lets\nothers use the modified version over a network has to offer them its source. A\nteam running an unmodified Grafana for its own staff stands somewhere very\ndifferent from a vendor shipping a patched one inside a product, and the vendor\nshould read the licence with counsel. Self-hosting is a single binary until a\nsecond instance is needed. From then on the instances share a MySQL or\nPostgreSQL database, and Grafana Cloud is the way to hand that work to Grafana\nLabs instead.\n\n## What it looks like in practice\n\nBoth times Grafana appears in this company's work, it is the part people looked\nat, with Prometheus behind it.\n\nIn an Infrastructure Engineer role from 2022 to 2024, the starting point was a\nteam that heard about problems after users did, because there was no single\nview of how its systems were doing.\n[A Grafana and Prometheus monitoring dashboard](/portfolio/cut-operational-risk-with-grafana-and-prometheus/)\ngave it that view. The key services were instrumented, the metrics were chosen\nbecause they meant something rather than because they looked busy, and alert\nthresholds were set on them and surfaced where the team would see them while\nthere was still time to act. Problems began to be caught before they escalated,\nreliability improved, and the team moved from firefighting to getting ahead of\ntrouble.\n\nThe other case was built for a mixed audience.\n[A dashboard on a large office TV](/portfolio/built-the-companys-first-observability-dashboard/)\nwas meant for technical and non-technical colleagues alike. Its second version,\non Grafana and Prometheus, showed metrics picked with the DevOps and engineering\nteams, fed by data pipelines that took in performance figures from several\nsystems, with alerts when a threshold was breached. Teams detected and resolved\nincidents 40% faster, and with the numbers in front of the whole office, the\nhealth of the systems became something people owned together.\n\nA wall screen like that is read from across a room by people who will never\nedit a query. It rewards a few large panels, units in the titles, and colours\nthat change only when someone should act, and Grafana leaves every one of those\nchoices to whoever builds the dashboard.\n\n## Where it goes wrong\n\n**A dashboard grows until nobody reads it.** Each incident adds a panel and\nnothing takes one away, so a year later the page holds forty small graphs and\nthe one that matters sits below the fold. Give each dashboard one audience and\none question, put the answer to \"is it healthy\" in the top row, and move detail\nto dashboards linked from there. A panel that nobody on the team can explain is\na panel to delete.\n\n**A fix made in the browser disappears.** The dashboard was provisioned from a\nfile. With `allowUiUpdates` at its default of false, Grafana refuses the save\nwith a \"Cannot save provisioned dashboard\" dialog. Set to true, it saves the\nchange to its database and then overwrites it with the file the next time the\nfile changes. Edit the file, or export the JSON from the save dialog and commit\nit, and keep the provider's setting explicit:\n\n```yaml\napiVersion: 1\nproviders:\n    - name: operations\n      type: file\n      allowUiUpdates: false\n      options:\n          path: /var/lib/grafana/dashboards\n```\n\n**The wall screen slows down the system it watches.** Every open browser, the\nTV included, re-runs every panel's query on each refresh, and five seconds is\nthe shortest refresh Grafana allows out of the box. Twenty panels on a\nfive-second refresh is four queries a second from one screen. Refresh at the\npace the data changes, which for Prometheus means the scrape interval or slower,\nraise the floor for everyone, and precompute heavy expressions with recording\nrules in the data source:\n\n```ini\n[dashboards]\nmin_refresh_interval = 30s\n```\n\n**The same incident pages people twice.** The rule lives in Prometheus, routed\nthrough Alertmanager, and again in Grafana Alerting. Or two Grafana instances\neach evaluate every rule, because alerting was left out when the second one was\nadded. Give every rule a single home, and follow Grafana's alerting\nhigh-availability steps before a second instance goes live.\n\n**Dashboards and users change depending on which server answers.** Two Grafana\ninstances behind a load balancer each kept their own SQLite file, so a dashboard\nsaved on one is missing on the other and a user exists on only one of them.\nGrafana's own guide makes a shared MySQL or PostgreSQL database a requirement\nfor high availability.\n\n## Further reading\n\n- [About Grafana](https://grafana.com/docs/grafana/latest/introduction/) — the\n  project's overview of what Grafana does, its editions and the open-source\n  projects around it.\n- [Data sources](https://grafana.com/docs/grafana/latest/datasources/) — the\n  systems Grafana can query, and how each one is added and configured.\n- [Dashboards](https://grafana.com/docs/grafana/latest/visualizations/dashboards/)\n  — panels, rows, variables and the JSON model a dashboard is stored as.\n- [Provision Grafana](https://grafana.com/docs/grafana/latest/administration/provisioning/)\n  — data sources and dashboards from files, and what happens to edits made in\n  the browser.\n- [Set up Grafana for high availability](https://grafana.com/docs/grafana/latest/setup-grafana/set-up-for-high-availability/)\n  — the shared database a second instance needs, and the extra steps for\n  alerting.\n- [Licensing](https://grafana.com/licensing/) — Grafana Labs' page on the 2021\n  move from Apache 2.0 to AGPLv3.\n\n\u003chttps://engineer.company/glossary/grafana/\u003e\n","date_modified":"2026-10-20T21:59:50Z","description":"Where Grafana gets its data, why its dashboards belong in version control, what AGPLv3 asks of you, and how one wall screen can slow the system it shows.","formats":{"gemtext":"https://engineer.company/glossary/grafana/index.gmi","gopher":"https://engineer.company/glossary/grafana/gophermap","html":"https://engineer.company/glossary/grafana/","json":"https://engineer.company/glossary/grafana/index.json","markdown":"https://engineer.company/glossary/grafana/index.md","pdf":"https://engineer.company/glossary/grafana/index.pdf"},"kind":"page","language":"en","title":"What is Grafana?","url":"https://engineer.company/glossary/grafana/","version":1}