Firing alerts now correlate into groups by shared labels, like service, host, or region. A single alert group opens per issue instead of paging separately on every individual signal.
How a group behaves:
- The first signal to fire sets the group's title, fixed after that even if a higher-severity signal joins later
- New matching signals append to the same group instead of spawning a new one
- Groups survive restarts and deduplicate against recurring issues over time
What you see inside a group:
- Every firing instance, with its state and severity
- A full audit trail, covering opened, appended, silenced, notification dispatched, and resolved states
- Per-channel delivery outcomes, with direct links to where each notification landed (Slack thread, Jira ticket, KloudMate Incidents)
Routing rules decide whether alerts group at all, along which label, and which channel the group notifies. Admins can also click Suggest rules to get AI-proposed grouping rules based on recent alert traffic, each with a preview of how much noise it would have collapsed.
A group is marked resolved once everything underneath has been Normal for a configurable period. A re-occurrence within the configured rejoin window rejoins the same group instead of opening a fresh one, so flapping alerts don't prematurely close and reopen the group.
This should cut a lot of noise during a real outage, when one root cause can otherwise trip dozens of alerts within seconds.
Check out the Alert Groups and Routing Rules documentation for full details.