Notify Groups: Routing and Escalation
Bundle channels once, attach the group to anything. Each channel sets its own severity floor, so a phone alert never fires for an informational notice.
A notify group bundles several notification channels together so you configure routing once and attach it to anything: a monitor, a business flow, a credential or a domain.
The alternative is picking channels individually on every resource, which works until you have forty monitors and need to change where alerts go. Then it is forty edits, and you will miss some.
Severity floors per channel
Each channel inside a group sets a minimum severity it wants to receive. That is what stops a phone call firing for an informational completion notice.
| Channel in group | min_level | Receives |
|---|---|---|
| info | Everything, including completions | |
| Slack | info | Everything |
| Slack (leadership) | critical | Only critical |
| SMS | critical | Only critical |
| Phone call | critical | Only critical |
The practical pattern is email and Slack at info so the team sees the full picture, and anything that interrupts a person set to critical. A single group can contain the same channel type twice with different floors, which is how you send everything to one Slack channel and only emergencies to another.
Severity comparison uses the underlying enum's ordinal order, so info, warning and critical rank as you would expect without any string parsing.
Escalation
An escalation timer on the group re-notifies when an alert has not been acknowledged. If nobody touches an alert within the window, it escalates and the group fires again.
| Free | Starter | Team | Enterprise | |
|---|---|---|---|---|
| Escalation | None | Fixed at 30 minutes | Configurable | Configurable |
| Notify groups | Yes | Yes | Yes | Yes |
| Channels per group | Email only | Email, Slack, Discord | All available | All available |
Starter's fixed 30 minutes is a deliberate simplification rather than a limitation for its own sake. Thirty minutes is a defensible default for a small team, and making it configurable is the kind of setting that matters once you have a rota.
Design your groups around who responds
The mistake is grouping by system. A group called production-alerts containing every channel means everyone gets everything, and within a month the channel is muted.
Group by responder instead. A payments group routing to the people who can fix payment problems, a data group for pipeline failures, an expiry group for renewal reminders that go to whoever handles renewals rather than to engineering at all.
That last group is worth building early. Executives asking to be added to alerting is common, and giving them a critical-only channel is far better than either refusing or adding them to everything.
Broken channels get flagged
Each channel records the status of its last delivery attempt. A channel failing repeatedly, such as a Slack webhook whose app was uninstalled, gets flagged automatically rather than failing silently.
This matters because a notification channel is itself a thing that can break quietly, and a group routing exclusively to a dead channel provides the appearance of coverage with none of the substance.
See notification channels for what each channel supports, and alerts for the lifecycle groups respond to.