Manage severity escalation notifications for the experimental alerting system

Use this page to control when a workflow runs as an alert episode's severity changes, including first-time matches, duplicate invocations, and de-escalation. Not every severity change invokes a workflow. The outcome depends on whether the action policy has already matched the alert episode and which frequency option you've selected.

Scope an action policy to the severity level you want a workflow to run for. When an alert episode escalates into that severity level for the first time, the action policy invokes a workflow because it has no prior invocation record for the alert episode.

The following example uses an action policy scoped to severity: "critical". An alert episode starts at low severity, so the action policy doesn't match. When the alert episode escalates to critical, the action policy now matches and invokes a workflow regardless of the frequency setting, because it has never invoked a workflow for this alert episode before.

Field Value
Match conditions severity: "critical"
Notify per Episode
Frequency On status change
Destinations PagerDuty workflow

If an action policy already matched an alert episode, a severity escalation alone doesn't trigger re-notification when the alert episode's status stays the same. With On status change frequency, a severity change doesn't count as a status change.

In this example, Action Policy A matches all alert episodes regardless of severity and notified when the alert episode was low. When the alert episode escalates to critical, Action Policy A still matches, but the throttle blocks re-notification because the status hasn't changed. To re-notify on escalation, use a time-based throttle or create separate action policies for each severity level as described in Route alert episodes by severity.

Field Value
Match conditions (None, matches all alert episodes)
Notify per Episode
Frequency On status change
Destinations Slack workflow

If an alert episode drops below an action policy's severity threshold, the action policy stops matching and doesn't invoke a workflow. If the alert episode later escalates back above the threshold, the action policy invokes a workflow again as if it were the first match.

In this example, Action Policy B targets only severity: "critical" alert episodes. An alert episode de-escalates from critical to high. Action Policy B no longer matches and stops invoking a workflow. If the alert episode later escalates back to critical, Action Policy B invokes a workflow again.

Field Value
Match conditions severity: "critical"
Notify per Episode
Frequency On status change
Destinations PagerDuty workflow