Skip to main content
A silence pauses notifications and escalation for matching incidents, while the incidents are still recorded. A semantic group collapses related incidents so that responders handle them once. Neither changes the rules.

​
Silence policies

Open Alerts → Silence policies. Each policy shows its state, its window, and what it matches. To create a policy:
  1. Select New silence policy and enter a Name. Keep Enabled on.
  2. Under Match, select Add matcher for each condition on the labels of incidents, such as service = checkout. All matchers must match. =~ and !~ take regular expressions. Without matchers, the policy matches Every incident.
  3. Under Window, choose the Window type:
    • Fixed range: a Start and an End;
    • Recurring: a Timezone, the Days of the week, and the hours from Hour start to Hour end (0 to 24), for example every Saturday from 2 to 6.
  4. Under Comment, enter the Reason, then select Create silence.
Silences match the labels that incidents carry: the labels of the rule, and the values of the groups of a split rule. Name the affected service and the end time in every planned silence, and let it end by itself rather than deleting it later.
A silence hides notifications; it does not prove that the service has recovered. Check the incident list when the window ends.

​
Silence one incident

Silence on an incident pauses notifications and escalation for that incident only, for 1 hour, 4 hours, 24 hours, or 7 days, with an optional Reason. It does not disable the rule or silence other incidents of the same service. The silence appears in Silence policies, where it can be changed or removed early. Creating, changing, and removing silences requires alerts.silence.

​
Semantic groups

Open Alerts → Groups. A semantic group collapses incidents that share label values, like the group_by of Alertmanager. Acknowledging or resolving a group through the API does the same to every incident in it; see Groups of incidents.
  1. Select New group and enter a Name, such as critical-by-service. Keep Enabled on.
  2. Under Matchers, add the conditions an incident must meet, for example severity = critical. All must match; with none, the group applies to every incident.
  3. Under Group by, enter the Group-by labels, comma-separated, such as service, region. Incidents with the same values of these labels collapse into one group.
  4. Select Create group.
A disabled group is skipped. Deleting a group keeps the incidents and removes only the grouping. Import creates groups from JSON: an array of groups, or an object with a groups array.
Managing groups requires alerts.manage.
Last modified on October 26, 2026