Skip to content

Rather more than uptime

Three things separate this from a checker that pings a URL: we confirm before we wake you, we understand what depends on what, and we stop escalating the moment the problem resolves itself.

Confirmation, not a single observation

One probe failing tells you about one probe, and most single failures are the network in between rather than your site. A failure is only an outage once a second probe on another continent agrees.

The states we use

Every state carries a shape as well as a colour. Hue never carries meaning on its own.

  • UP Healthy. Answering as expected.
  • CLEAR On no blocklist we check.
  • VALID The certificate checks out.
  • DOWN Confirmed from a second continent.
  • EXPIRING A threshold crossed. Not an outage.
  • PAUSED Deliberately not checked.
  • NO DATA Our probe is unreachable. We know nothing, and we say so.

An unreachable probe is never treated as an outage. If we cannot see, we do not guess. That is what NO DATA means, and it is never an alert.

What depends on what

One alert, not thirty

Tell us a site depends on its server. When the server goes down you get one message about the server, not one for every site on it. Suppression and grouping stay separate concepts: a group rolls status up, a dependency stops noise going out.

Escalation that cancels itself

Text yourself immediately, Slack the team after two minutes, email the owner after forty-five. If it recovers first, the later messages never go out: they are cancelled, not sent and retracted.

What we watch

Your mail server going quiet or your IP landing on a blocklist costs you as much as a down homepage, and usually takes longer to notice.

Want it when it opens?

One message when it launches. Nothing else, ever.

Create an account