Since you’re on Alertmanager, the nice trick is routing by severity: keep a webhook_configs receiver for the noisy stuff (ntfy is great for that) and add a second receiver only for severity=critical that goes somewhere able to break through Do Not Disturb, so disk-full warnings don’t wake you but a dead NAS does. That also keeps you off your own infra like Discord did, as long as the critical path is a hosted service. Full disclosure, I’m the dev of Echobell, which turns a webhook into a push, a time-sensitive alert or an actual phone-call-style alert per channel, so it slots into that second receiver; not FOSS though, so if self-hosting the notifier matters, ntfy + Pushover’s emergency priority is the open-ish combo.
Since you’re on Alertmanager, the nice trick is routing by severity: keep a webhook_configs receiver for the noisy stuff (ntfy is great for that) and add a second receiver only for severity=critical that goes somewhere able to break through Do Not Disturb, so disk-full warnings don’t wake you but a dead NAS does. That also keeps you off your own infra like Discord did, as long as the critical path is a hosted service. Full disclosure, I’m the dev of Echobell, which turns a webhook into a push, a time-sensitive alert or an actual phone-call-style alert per channel, so it slots into that second receiver; not FOSS though, so if self-hosting the notifier matters, ntfy + Pushover’s emergency priority is the open-ish combo.