Changelog ·

Alert mode: outages confirmed in under a minute

When a probe sees a failure, the whole fleet now checks your site every 5 seconds, confirms a full outage in about 47 seconds instead of 73, and alerts you when the site is down where your visitors are.

alertschecks

Since 20:53 UTC every account is on Alert mode. Until now a failure seen by a scheduled round started confirmation rounds 20 seconds apart, and a site was marked down after two of them, about 73 seconds after the first failed check at the median. Now every healthy probe also joins in at once, in three groups that take turns, so your site is asked every 5 seconds, step by step: DNS, connection, TLS, HTTP.

  • Faster, with the same rule. A site is down when two thirds of the healthy probes, on three continents and three providers, agree for 20 seconds. In two weeks of testing on 439 monitors that came a median of 47 seconds after the first failed check; outage by outage, a median of 22 seconds sooner than before. The alert says where the request broke.
  • Down where your visitors are. If most of our probes on one of the monitor's audience continents can't reach the site for a minute, you get «Down for your visitors in Europe» even while the rest of the world still reaches it, and the time counts as downtime. We pick it for you, usually the continent nearest your server, and you can change it on the monitor.
  • Short blips stay quiet. A blip that clears inside 20 seconds, or a regional one under a minute, sends nothing, and a site that fails where it always fails (a geo-restricted site) starts nothing new.

Alert mode is free, like confirmation rounds, and changes nothing in your settings, channels or credits. Details in How a check works.

← All updates