Alerts
An alert rule watches one metric against one target, a host or a guest, for a threshold breach sustained over a duration you choose, not a single reading that happens to cross the line. Proxploy walks a target’s recent samples backward from now and only counts a breach if the threshold has held continuously for at least that duration; one healthy sample in the middle resets the clock, so a brief dip below threshold doesn’t quietly keep an old breach alive. Rules are checked at creation time against what a metric can honestly be measured for: a disk-usage rule against a guest is rejected, because that number is only meaningful for a host.
Firing and recovery
Section titled “Firing and recovery”At most one alert is open per rule-and-target pair at a time. Once a breach resolves, Proxploy closes the alert automatically; there’s no separate “mark resolved” step for a metric that’s back in range.
Getting notified
Section titled “Getting notified”Firing and resolving alerts publish live to the UI immediately, and separately can notify through configured channels. If a notification fails to send, it never costs you the live update: the UI event always goes out first, and notification delivery is best-effort on top of it.
Acknowledging
Section titled “Acknowledging”You can acknowledge a firing alert without resolving it, useful for “I’ve seen this and I’m working on it” without waiting for the metric itself to recover.