Alerts only when several cities agree

One city is not enough

A check from one place fails for reasons that have nothing to do with your site: a congested link at the checking provider, a routing problem between two networks, a resolver that times out once. Alert on every such failure and you learn to ignore alerts. Wait for many failures from the same place and a real outage reaches you late.

BBoxProbe asks the question from several cities instead. When your site is down, it is down from everywhere. When only one city fails, the problem is usually on the way to your site rather than at it.

The rule: most of a monitor's cities

A round counts as down when enough of the monitor's cities fail in that round. Unless you set the number yourself, enough means more than half of its cities, and at least two when it checks from two or more. Rounds are aligned to the clock, so in each round every city's check starts at the same time.

You can set your own number in the monitor's settings, under Cities that must fail: anything from 1 up to the number of cities you picked. Set it to 1 and any city failing counts; set it to all of them and only a failure everywhere does.

Fewer failing cities than that is not nothing: the monitor shows as partly down, with the failing cities and what each one saw, and a status page that shows it says "Some cities see errors". You just don't get an alert for it.

Two rounds in a row

A single bad round can be a blip. So by default a city counts as failed only when its newest check and the one a round before both failed, and the monitor is down when enough cities failed that way. In the settings, Confirmation rounds is 2 by default; set it to 1 to hear about the first failed round.

A round is one check from each of the monitor's cities. How often rounds come is up to you, as often as your plan allows: every 2 minutes on Free, every minute on Pro, and every 30 seconds on Max. With two rounds, an outage is reported after its second failed round. Each city waits for an answer as long as the monitor's Timeout: up to 3 seconds on Free, 4 seconds on Pro, and 5 seconds on Max, the most the plan allows unless you set it shorter.

When it is back

With two confirmation rounds, the alert ends one interval after the first good round: a lone failed round in that time doesn't hold it open, and opening it again takes two failed rounds again. With one, it ends at the first good round. If it goes down and up three times or more in two hours, you get one "unstable" alert instead of a notice for each change; How it works has the details.

What the alert tells you

The alert names the cities that fail and what each one got, so you can tell a server error from a timeout or a failed text check at a glance. It is the same in every channel: email, Slack, Discord or your own webhook.

Picking cities

A new monitor starts with Frankfurt, Virginia, and Singapore: one each in Europe, North America and Asia, so a regional outage shows as one city failing rather than all of them. On a plan with fewer cities per monitor than that, it starts with the first ones. Change them whenever you like.

The cities

The checks run from 15 cities on 4 continents. Every plan picks from all of them.