How monitoring checks and failure confirmation work

Learn how Uptime Tracker schedules checks, confirms failures before alerting, decides when a monitor recovers, and why some results are shown as unknown.

Updated By the Uptime Tracker team

Uptime Tracker runs each monitor on its own schedule, confirms any failure before calling it an outage, and only then opens an incident and sends alerts. A single failed check never pages you on its own.

Check intervals

Each monitor has a Check interval (seconds) between 15 and 3,600 seconds, limited by your plan's fastest interval:

PlanFastest interval
Free300 seconds (5 minutes)
Pro60 seconds
Team30 seconds

Each check also has a Timeout (seconds) between 1 and 30 seconds, at least 2 seconds shorter than the interval. Domain-expiry monitors are the exception: they run hourly from a cached registry answer.

How a failure is confirmed

When a check fails, the probe re-checks the target quickly before deciding anything. A monitor is marked down only after two consecutive failures. This filters out one-off network blips and short restarts that would otherwise wake you up for nothing.

Monitoring currently runs from one probe location. Additional probe regions are planned; once they exist, confirmations from other regions will also be required before a monitor is marked down.

How recovery works

A monitor that is down needs two consecutive successful checks before it counts as recovered. When that happens, the incident is resolved and a recovery notification goes to the same destinations.

What counts as a failure

  • HTTP(S): a connection error (such as a refused or reset connection), a timeout, an unexpected status code (by default anything outside 2xx/3xx), a missing keyword, an invalid, expired, untrusted or wrong-hostname TLS certificate, or a domain that no longer exists in DNS. Redirects are followed, up to 5 hops.
  • TCP, ping, gRPC, SMTP, IMAP: the connection or handshake does not succeed within the timeout.
  • DNS: the record does not resolve, or no answer contains the expected value.
  • Heartbeat: no ping arrives within the expected interval plus the grace period.

Unknown results and platform errors

If a result is uncertain, for example because of a problem on our side rather than yours, the monitor shows unknown instead of a fake up or down. Probe and platform errors are not counted as your downtime, so they do not lower your availability figures.

When a domain stops resolving

For HTTP(S) monitors, DNS answers are treated by what they say about your site:

  • If DNS answers definitively that the name does not exist (NXDOMAIN) or has no addresses, for example because the domain expired or the record was deleted, the check fails with DNS: no such host. After the usual confirmation the monitor is marked down and you are alerted.
  • A DNS timeout or a resolver failure (such as SERVFAIL) says nothing certain about your site, so it is recorded as unknown and never alerts on its own.

How availability is calculated

Availability is time-weighted per monitor. Outages are backdated to the first failed check, not the moment confirmation finished, so the figures reflect real impact. Time inside a maintenance window is excluded from availability. Paused monitors do not count toward uptime or plan slots.

What an alert contains

Alerts include the monitor name, when the outage was confirmed, a Reason: line saying why the check failed, and a redacted target (for HTTP monitors, only the host is shown), so a message never leaks tokens from a URL's query string. The reason is a short summary, never a response body, header or URL. Typical reasons:

What happenedReason shown
Unexpected status codeHTTP 503 Service Unavailable
No answer within the timeoutTimed out after 10s
Certificate problemTLS: certificate expired or not yet valid
Server not accepting connectionsConnection failed: connection refused
Keyword missing from the pageKeyword not found (HTTP 200)
Domain no longer existsDNS: no such host
Heartbeat not receivedNo heartbeat received within 10m (expected every 5m plus 5m grace)

The reason is recorded when the incident opens. It appears in down alerts, repeat reminders and escalation pages on every channel (email, chat, push, SMS and PagerDuty), as failure_reason in webhook events, and as Reason on the incident page. The latest check results, including later ones, are listed on the monitor's Check results page.

FAQ

Frequently asked questions

Will one failed check send me an alert?

No. A monitor is marked down only after two consecutive failures, and only then are an incident and alerts created.

Do you check from multiple regions?

Not yet. Monitoring currently runs from one probe location, and additional probe regions are planned.

Why does my monitor show unknown?

Unknown means the result was uncertain, for example because of a probe or platform error or a DNS timeout. These periods are not counted as your downtime.

Does the alert say why my site is down?

Yes. Down alerts, reminders and escalation pages include a Reason line, such as HTTP 503 Service Unavailable, Timed out after 10s or DNS: no such host. The incident page shows the same reason.

Is an expired domain reported as down?

Yes. When DNS answers that an HTTP monitor's domain does not exist or has no addresses, the check fails with DNS: no such host and alerts after the usual confirmation. DNS timeouts and resolver failures are shown as unknown instead.

Uptime Tracker

Start monitoring in under five minutes

Start on the free plan — commercial use allowed. No credit card required.

  • Free forever plan
  • No credit card
  • Cancel anytime