Skip to content

← All news

A slow handshake no longer turns a certificate monitor red

A certificate is checked twice a day, so one slow handshake is the whole of that day's evidence. Until now a single timeout opened an incident and turned the monitor red, and the next look at it was the recheck of a monitor that was already down. One blip on a busy afternoon read as an outage.

What changed

  • The check tries the handshake three times, two seconds apart, before it reports a timeout, a refused connection or a TLS error. A host that answers is never delayed by this, and a host that does not is reported after about a minute.
  • If it still cannot reach the host, the monitor shows CHECKING rather than DOWN, and looks again in five minutes.
  • Only the third failure in a row opens an incident, about ten minutes after the first. A good check in between starts the count again.
  • A certificate that is expired, or that we could read but not make sense of, is the same every time and still opens an incident straight away. Waiting would only delay a true answer.

Unreachable is not invalid

The monitor page used to call a certificate we could not reach INVALID, which sent people to check an expiry date that was months away. A timeout, a refusal or a TLS failure now reads UNREACHABLE, the same word the monitors list already used, and INVALID is kept for a certificate we actually read and found wrong.

An incident that was already open when this shipped keeps its old wording until it resolves.

Read more