Alerting your team when e-mail is not enough

The need

Your monitoring already sends e-mail. E-mail arrives in a mailbox somebody has to be looking at, behind a connection that may be the very thing that broke. At three in the morning, a phone that rings is the only channel that works.

The usual answer is a paging service with a monthly fee per responder. If your alerts are few and your team is small, that is a subscription for a handful of messages a month.

The setup

1. One call from your monitoring. Prometheus Alertmanager, a cron, a health check in your own code: anything that can make an HTTP call can raise an alert. One endpoint, one key.
2. A routing rule by prefix. Send the on-call numbers through the SIM whose carrier serves them best, so the alert does not depend on a route that is slow at night.
3. A reply that acknowledges. The responder answers "ok" and the reply lands in Conversations, on their thread. A rule can label it, or call a webhook to close the incident in your own tool.
4. The gateway watches itself. It raises its own alerts, by SMS, when a phone drops, a battery falls low or a SIM reaches its daily quota. Those never pass through anyone else either.

What to watch

Keep an alerting SIM away from campaigns. A daily quota spent by a marketing send is a quota that is no longer there for the alert at midnight. Take the phone out of the routing pool, or give it a quota of its own.

An alert is transactional traffic. Mark it as such on the send, and the send window and the night pause will never hold it: they hold campaigns, and only campaigns.

Your gateway is now part of the chain you monitor. It runs on a machine you own, which means you watch it like the rest: the uptime history of each phone is on the Devices screen, and the alerts tell you before a silent phone costs you an incident.