False positives

A false positive is an alert for an outage that never happened: your service was up, but the check said otherwise. Hyperping confirms every failure from multiple regions before it opens an outage, and this page covers the rest of what you can tune.

Most fake alerts come from one of three places:

  • A network blip between one probe and your server, not an outage on your side
  • A monitor configured more strictly than your service actually behaves
  • Something between you and the internet, like a WAF or a rate limiter, blocking the probe

How a failure gets confirmed

Checks rotate through your selected regions, one region at a time. When one of them fails, the failure is not trusted on its own:

       check from London
               |
               v
            FAILURE
               |
               v
   +-----------------------+
   |  instant double check |
   |  from your other      |
   |  regions              |
   +-----------------------+
               |
       +-------+-------+
       |               |
    all fail      any succeeds
       |               |
       v               v
 outage opens      no outage
 alerts sent       no alert
               kept in the logs
  1. One region reports a failure

    Bad status code, a broken assertion, an unreachable host, or a timeout. Nothing is sent to your team yet.

  2. Every other selected region re-checks immediately

    The double check runs at once, not on the next interval, so a confirmed outage still reaches you in seconds.

  3. Only a confirmed failure opens an outage

    If the other regions answer normally, no outage opens and no alert goes out. The failed check stays visible in the monitor's logs.

The probes also run on a mix of hosting providers (AWS, Scaleway, DigitalOcean) across 18 regions, so a routing problem at one provider cannot fail every check at once.

Settings that cut the remaining noise

SettingWhat it prevents
Alert delay (alerts_wait)Alerts on blips that resolve on their own. With a 5 minute delay, a 2 minute outage never pages anyone.
Expected status codeDown alerts on endpoints that legitimately answer 401, 403 or 404.
Follow redirectsDown alerts on a 301 or 302 that is the intended response.
RegionsSingle-path failures. More regions mean more independent confirmations.
Maintenance windowsAlerts during a planned deploy or migration.
Grouped alertsTen notifications for one widespread incident.

The alert delay lives in the monitor's Notifications tab, or in the alerts_wait field of the update monitor API, in minutes. Everything else is in the monitor settings.

When an alert still looks wrong

Open the monitor's logs first. They show the status code, the response time and the region for every check, which usually points straight at the cause.

The service works in my browser but the check fails

Your browser is not the probe. Look for a status code outside your expected range, a redirect that is not followed, or a text assertion that no longer matches the page. Logs show what Hyperping actually received.

Checks get 403 or 429 responses

A WAF, a bot filter or a rate limiter is blocking the probes. Datacenter IPs trigger these rules more often than residential traffic. Allow the probe regions in your firewall, or raise the rate limit for them, and lengthen the check interval if the endpoint is expensive.

A keyword assertion fails after a copy change

Text assertions break when the page text changes. Assert on something stable, like an element that only renders on a healthy page, rather than marketing copy. See keyword monitoring.

The alert fires every time we deploy

Schedule a maintenance window for planned work, or pause the monitor for the duration of the migration. For rolling deploys that drop a few requests, an alert delay of 2 to 3 minutes is usually enough.

A browser check fails but the flow works manually

Browser checks fail on any broken assertion or on the 2-minute run timeout. Open the run's screenshot, video and Playwright trace to see where the script stopped. Failed runs are already retried from a different region before an outage opens. See browser checks.

Only one region ever fails

That is a regional problem, not a false positive, and it is worth investigating: a CDN edge, a DNS record or a peering route can be broken for part of the world while the rest is fine. Compare response times per region in the logs.

Next steps