Alerts only for a real outage

Because one failed check is not an outage. The internet drops a request now and then, between one of our locations and your server, while your visitors notice nothing. So when a check fails, You Up? tries again, then asks the Monitor’s other locations to look. You hear about it only when the failure is confirmed from other locations: a real outage, not a blip.

What happens when a check fails

1. We try again. The location that saw the failure tries once more, a few seconds later. When that gets through, it was a blip, and you hear nothing.
2. We ask the other locations. When the second try fails too, the Monitor’s other locations look at once.
3. Then we tell you. When the failure is confirmed from other locations, an Incident opens and you get one alert, saying where the request stopped.

The same goes the other way: you hear that the site is back only once the other locations get through too.

Free watches from one location

Free watches from one location, so it has no other location to ask. It still tries again before it tells you, so a single dropped request does not wake you, but a problem on the way between that one location and your server can. Every paid Plan watches from five locations, and confirms each failure from the others.

When only one location cannot reach you

Sometimes one location cannot reach your site while the others can. For everyone the others stand for, your site is up, so no Incident opens and nothing counts against your uptime. If it lasts, you hear about it as a Regional outage: worth knowing, because people near that location may not reach you either.

One cause, one message

When a server stops answering, its certificate cannot be read either, and several Monitors can fail in the same minute. You still get one message, which says what went wrong, not one for each thing that failed with it.

How a failure is confirmed, in the Docs

Questions

Can You Up? still get it wrong?

Rarely, and when it does, we say so. If a location that confirmed a failure is later found to be at fault itself, we take the Incident back: it is marked as reported in error, and it no longer counts against your uptime or shows as downtime on your Status page.

Does checking again make alerts slower?

By seconds, not minutes. The second try comes a few seconds after the first failure, and we wait at most half a minute for the other locations to answer.

What about heartbeats?

A Heartbeat Monitor has no locations: your job calls us, so there is nothing to confirm from elsewhere. The alert goes out when a call is late.

What if too few locations can answer?

Then we decide nothing and send nothing, rather than guess. The time counts neither for nor against your uptime, because the fault is ours.