Tech Antenna logo

9 Reasons Grafana Is the Most Important Tool for Metrics Tracking (Including the Time It Showed Up at SpaceX)

From a Falcon 9 launch control room to 10,000 paying customers, here is why I keep coming back to Grafana for every metrics stack I build.

By Jordan Lee
9 Reasons Grafana Is the Most Important Tool for Metrics Tracking (Including the Time It Showed Up at SpaceX)

I have set up a lot of monitoring stacks over the years. Prometheus for scraping, Loki for logs, InfluxDB for time series, CloudWatch because AWS made me, and a rotating cast of paid platforms that promised to replace all of it. The one constant across every single stack, in every job, at every company size, has been Grafana.

That is not brand loyalty talking. It is the simple observation that Grafana sits at the layer where humans actually make decisions. Your database stores the numbers. Your agent collects them. But somebody has to look at a screen at 3 a.m. and decide whether to roll back the deploy. That screen is almost always Grafana.

Below are the nine reasons I consider it the single most important tool in the metrics tracking category, with the SpaceX story woven through because it is genuinely the best illustration I know of what this tool is for.

1. The SpaceX moment proved Grafana belongs in high-stakes rooms

In January 2016, a Grafana Labs engineer named Dieter Plaetinck was watching a Falcon 9 launch video and noticed something familiar on one of the screens in the background. It was a Grafana dashboard, running inside the SpaceX Launch Control Center.

The company wrote about it the same week in a post titled Grafana: It's Rocket Science. What I like about that post is how unpolished the reaction was. The team described it as "a thrill to find out when someone cool is finding value in your software", and admitted they had emailed SpaceX to find out more, which tells you they had no idea it was happening.

Think about what that means in practice. A company that lands orbital-class boosters on drone ships, where a bad decision costs tens of millions of dollars and years of schedule, chose an open source dashboarding tool to keep eyes on critical systems. Nobody sold it to them. An engineer installed it because it worked.

Grafana co-founder Torkel Ödegaard has referenced the moment repeatedly since. In a retrospective marking Grafana 10, he listed the SpaceX control center among the places he had spotted his own software and said "realizing that Grafana is used by over 20 million users, just blows my mind."

2. SpaceX did not stop at rockets, and neither did the community

The second half of the SpaceX story is Starlink, and it is arguably more interesting because you can reproduce it yourself this afternoon.

Every Starlink dish exposes a gRPC endpoint full of telemetry. The community built exporters around it, and the reference project, danopstech/starlink, ships Prometheus, Grafana, a Starlink exporter, a speedtest exporter, and a blackbox exporter in a single Docker Compose file. It "collects information from the Starlink dish every 3 seconds", covering signal strength, alarms, obstructions, and latency.

The metric I find most charming is obstruction tracking. The exporter breaks the sky above your dish into twelve 30 degree wedges and reports the obstructed fraction of each one. That is a physical, spatial problem (a tree branch in the wrong place) translated into a time series you can graph, alert on, and act on. That is the entire thesis of metrics tracking in one example.

So SpaceX appears at both ends of the scale: institutional Grafana inside a launch control room, and grassroots Grafana on a Raspberry Pi in somebody's garage, watching a dish talk to satellites.

3. It reads your data where it already lives

This is the structural reason Grafana wins, and it took me a few years to appreciate it fully.

Most monitoring vendors want to be your database. Grafana does not care. It queries Prometheus, Loki, Elasticsearch, InfluxDB, MySQL, PostgreSQL, Snowflake, CloudWatch, Azure Monitor, Google Cloud Monitoring, Jira, and dozens more. Grafana's own monitoring integrations catalog lists everything from Kafka and Cassandra to Home Assistant and Raspberry Pi in a single browsable index.

Analysts at Sacra describe this as a "big tent" philosophy that supports over 100 databases and data sources, including competitors' tools. The practical effect is that adopting Grafana costs you nothing architecturally. You are not migrating. You are pointing a viewer at storage you already pay for.

Compare that to the standard SaaS pitch, which requires you to ship all telemetry into a proprietary store before you see a single graph.

4. The adoption numbers are no longer a hobbyist story

I used to hear "Grafana is fine for startups" as a serious objection. That argument has aged badly.

Metric Figure Period
Active Grafana instances 1,000,000+ Nov 2022
Total users 20M+ (some estimates now higher) 2023 onward
Paying customers 10,000+ Aug 2026
Annual recurring revenue $600M+ Aug 2026
Grafana Cloud monthly active users ~127,000 to 251,000+ 2-year growth
Grafana Cloud signups 500,000+ Past 2 years

Those last figures come from Grafana Labs' August 2026 milestone announcement. The same release noted that self-serve monthly active organizations grew from roughly 42,000 to 66,000, that more than 1.1 million distinct people logged in during that window, and that over 18,000 organizations are actively using Grafana Assistant.

Wikipedia's entry on Grafana currently lists Nvidia, Anthropic, and Uber among its paying customers. Grafana Labs has also been named a Leader in the 2026 Gartner Magic Quadrant for Observability Platforms for the third consecutive year.

5. It solves the problem the industry says is worst

Grafana Labs runs an annual Observability Survey, and the 2026 edition is the largest yet. It drew 1,363 responses from engineers, SREs, and technology leaders across 76 countries, collected between October 2025 and January 2026.

The findings are blunt. Complexity and overhead topped the list of concerns at 38%, ahead of signal-to-noise problems at 34% and cost at 31%. Alert fatigue was the single biggest obstacle to faster incident response at 30%, nearly double the next answer.

That is exactly the pain a unified dashboard layer addresses. The same survey found that more than three-quarters of respondents, 77%, say they have saved time or money through centralized observability.

We have felt this ourselves. Before consolidating, my team checked four separate consoles during an incident. After consolidating on Grafana, we checked one, and mean time to identification dropped noticeably without us changing a single line of application code.

6. Dashboards as code, which matters more than it sounds

Grafana dashboards are JSON. That means version control, code review, and CI/CD. Combined with the Terraform provider and Grafana's provisioning system, your dashboards and data source configuration become reproducible artifacts rather than tribal knowledge trapped in one person's browser.

I treat this as a hard requirement now. If a monitoring tool cannot express its own configuration as text I can commit, it does not go into production.

7. Alerting is unified across the whole stack

Grafana Alerting has consolidated into a single engine covering the entire stack. At GrafanaCON 2026, the team described Grafana Alerting supporting more than 50 data sources and 22 integration types, with a new active/standby high availability mode in Grafana 13 and an import tool that brings existing Prometheus, Mimir, and Loki alert rules across without copy-paste.

One alerting engine, many backends. That is the difference between an on-call rotation that works and one that quietly loses pages.

8. Grafana vs the usual alternatives

Here is how I actually reason about the choice.

Grafana vs Datadog. Datadog is excellent and expensive, and it owns your data. Grafana is cheaper, open, and asks you to assemble a bit more. If cost predictability at scale matters, and the 2026 survey says cost is still a top selection criterion, Grafana usually wins. If you want zero assembly and have budget, Datadog is defensible.

Grafana vs Kibana and Elasticsearch. Kibana is superb for log search and weaker for time series visualization across mixed sources. Grafana queries Elasticsearch happily, so this is often not either/or. I run both and use each for what it is good at.

Grafana vs Prometheus alone. This is not a real comparison. Prometheus is the database and Grafana is the face. Prometheus ships a basic expression browser that nobody wants to stare at during an outage.

Grafana vs building your own. I have watched teams try. They spend two quarters rebuilding a worse version of a panel library that already has over a hundred visualization types.

9. It scales down as gracefully as it scales up

The best argument for Grafana is that the same tool serves a solo developer and a national space program.

Firefly Aerospace presented at GrafanaCON about running Grafana for their Blue Ghost lunar lander mission, which launched on a SpaceX Falcon 9. The moment their engineer described that stuck with me was the mission control room watching dashboards after separation, when the panels "said no data." Every operator on earth knows that feeling.

At the other end, a university rocketry team used Grafana to break a world record, monitoring tank pressures, temperatures, and valve states from a bunker in the Mojave. They noted that the dashboard let them "view live data from the tanks and quickly notice issues", and caught a misbehaving relay right before the record launch because of it.

Torkel described his original 2013 intent simply: "my goal was to make time series data accessible for a wider audience." A decade later, that goal is the reason the same tool works for a hobbyist dish and a lunar lander.

Where I would push back on my own argument

I do not think Grafana is flawless, and pretending otherwise would be dishonest.

Self-hosting has real operational cost. Somebody owns upgrades, plugin compatibility, and the backing stores. Dashboard sprawl is a genuine problem, and I have inherited instances with 400 dashboards where nobody knew which twelve mattered. And Grafana is a visualization and alerting layer, not a data pipeline, so a bad instrumentation strategy underneath will produce beautiful, useless graphs.

None of that changes the conclusion. If I could keep exactly one tool in my metrics stack, I would keep the one that reads from everything else, costs nothing to start, runs on a Raspberry Pi, and has been photographed on the wall of a rocket launch control room.

Sources