Alerting

Collecting signals is useless if no one is watching them at 3 a.m. Alerting is the active half of the watch: thresholds and rules that, when crossed, push a notification to a human instead of waiting to be found. A disk filling, a service down, a security event that shouldn’t be — each has a line, and crossing it raises a hand.

What triggers an alert, and how it reaches someone, is defined here. The collection it draws on is Signals; the outside liveness check that fires when everything else can’t is Uptime.

Has anything touched?

If reading this made you want to argue with it, extend it, or notice what's missing, that's the signal to show up.

:/back-to-top