Skip to content
AlertPing

Fly.io uptime monitoring with outside checks and downtime alerts every 30 seconds

Short answer: Fly.io health checks decide whether the Fly Proxy sends traffic to a Machine. They do not notify you, and Fly states it has no built-in alerting on metrics. To know within seconds that your public app stopped answering customers, you need a check that runs from outside Fly.io and pages a person.

Check a URL first ▸

Fly.io behavior checked against docs.fly.io

alertping ▸ check preview

simulated

▸ type a domain to watch a check and an alert play out

▸ simulation of 3 regions · US East · EU Central · EU East

▸ waiting…▌

If ever goes down, you get:

Alert fired ▸ 2 channels · 6.2 s after first failure

AlertPing app

● DOWN : HTTP timeout confirmed from 3/3 regions (US East, EU Central, EU East). Incident opened.

sms · on-call

AlertPing: DOWN. Confirmed 3/3 regions . First fail: EU Central.

Simulation, not a measurement of your site: this preview runs entirely in your browser and the latencies are generated from the domain you typed. Real checks on your monitors run from our US East, EU Central and EU East regions.

read this first

A Fly.io health check is a routing decision, not an alert

Fly.io documents its health checks clearly. Service-level TCP and HTTP checks confirm that your app listens on a port or returns a 2xx response on a path. When one fails, the Fly Proxy marks that Machine unhealthy and stops routing requests to it until the check passes again. The docs add that a failing check will not restart or stop the Machine, and top-level [checks] do not affect routing at all.

That is useful plumbing and it is not monitoring. Nothing in a health check emails, texts or pages you. The health checks page points you to fly checks list to see status, and Fly's metrics documentation says Fly.io does not include built-in alerting on metrics, so you set up alerting yourself against the Prometheus endpoint. In practice the first person to notice an outage on a small Fly app is a customer.

There is a second problem. A check that runs inside Fly's network, against a port, can pass while the public path is broken: a bad certificate on your custom domain, a DNS change, a 301 redirect that your HTTP check refuses to follow, or every Machine marked unhealthy so the proxy has nowhere to send traffic. Customers see the result of all of those, and only a request from outside sees it the way they do.

the honest table

What Fly.io catches and what only an outside check catches

FailureFly.io on its ownAlertPing from outside
An app crashes and the Machine stopsExit metrics exist, but no alert unless you build oneYes, the public URL fails and you are paged
A failing health check pulls a Machine out of routingRouting changes silently, no notificationYes, if the remaining capacity errors or slows
Every Machine is unhealthy and the proxy has nowhere to routeVisible in fly checks list if you lookYes, within 30 seconds, confirmed from a second region
Custom domain DNS breaks or the domain lapsesNoYes
TLS certificate problem on a custom domainNoYes, with days of warning before expiry
The app returns 200 with an error page or empty bodyHTTP checks accept any 2xxYes, with a response string assertion
A Postgres or API dependency fails and requests return 500Only if your check path touches itYes, check the route that uses it
An app with no public service (a worker)Top-level checks, no alertNot its job, use a heartbeat or watchdog

Keep the Fly.io health checks, they keep traffic off broken Machines. Add an outside check for the question customers care about, which is whether the app answers at your own domain. Fly also publishes its health check reference if you want to tune the grace period and interval.

the mechanics

Setting up outside monitoring for a Fly.io app

Step one, point the check at your custom domain rather than the fly.dev hostname, so DNS and certificate failures are in scope. Step two, add a check for each route that earns money: the login or checkout path, and any API route the front end needs. If your app forces HTTPS and your own health check fails on the redirect, the outside check follows it the way a browser does.

Step three, assert on a string in the response and on response time, not only the status code. A Machine that serves a framework error page still returns 200 to a check that looks only at the status. Step four, choose 30-second checks for anything customer-facing, confirmed from a second region before it counts as downtime, so one network blip does not wake anyone. Step five, route alerts to SMS and Slack. SMS, email, Slack and webhook alerts are included on every AlertPing plan with no credit packs. Nothing is installed in your Fly.io organization and no fly.toml changes are needed.

If you run in several regions, one check against the public hostname tests whatever the Fly Proxy routes a user to. Add separate checks against region-specific hostnames only if you expose them.

who needs this

Which Fly.io teams need this

Startups that put a paying product on Fly.io Machines and have no on-call tooling yet. Agencies running client apps on Fly.io, where an outage becomes a phone call. Teams that deployed to Fly.io for low latency in several regions and now have customers expecting it to stay up. If one missed outage costs you more than a monitoring plan, you are the reader this page is for.

scope

What AlertPing does not replace on Fly.io

AlertPing does not read your Machine logs, Prometheus metrics, CPU or memory, and it does not replace Fly's health checks for routing or the managed Grafana dashboards. It watches the public side: the URLs your customers and integrations call, how fast they answer and whether the certificate is healthy. The same gap shows up on Heroku, where alerting depends on the dyno type.

questions people ask

Fly.io uptime monitoring questions

Does Fly.io have uptime monitoring?

Fly.io has health checks that control routing and managed Prometheus and Grafana for metrics. Its documentation says there is no built-in alerting on metrics. None of it requests your custom domain from outside on a schedule and pages you when real requests fail.

How do I get alerted when my Fly.io app goes down?

Point an external monitor at your custom domain, add checks for the routes that matter, assert on a known response string, confirm failures from more than one region and send the alert to SMS or Slack. AlertPing does this every 30 seconds without changes to your Fly.io setup.

What happens when a Fly.io health check fails?

The Fly Proxy marks the Machine unhealthy and stops routing requests to it until its checks pass again. Per the docs, the Machine is not restarted or stopped automatically, and no notification is sent.

Can AlertPing monitor an app with several Fly.io regions?

Yes. A check against your public hostname tests whatever the Fly Proxy routes a user to. If you expose separate hostnames per region, add a check for each one.

Does a passing Fly.io health check mean my app is up?

No. It means the Machine answered the check Fly runs, on a port or path you chose. It does not test your custom domain, certificate, redirects or the full request path a customer takes.

Watch your Fly.io app from outside Fly.io

AlertPing checks your custom domains and API routes as often as every 30 seconds, confirms every failure from a second region, warns before your SSL certificate expires, and sends SMS, email, Slack and webhook alerts on every plan. Flat pricing from $49 a month, with 30-second checks from $59.

See pricing