Loading

Review action policy execution history in the experimental alerting system

Action policy execution history shows dispatcher decisions from the last 24 hours across all action policies in the space, so you can confirm notifications are dispatching as expected or investigate unexpected notification behavior.

Go to Execution history in the navigation menu or global search, then select the Policies tab. Each row covers one dispatcher run for each action policy evaluated against a rule:

Column Description
Timestamp When the dispatcher ran.
Policy The action policy that was evaluated.
Outcome Whether the dispatcher acted on the episode: dispatched, throttled, or unmatched. Definitions are in Dispatch outcomes.
Rules The rule whose alert episodes the action policy processed.
Episodes The number of alert episodes processed in this run.
Action groups The number of action groups involved.
Workflows The workflows invoked, if any.

You can search records by action policy name, rule name, or saved-object ID, and filter by outcome to view only dispatched or throttled records.

After each dispatcher run, Kibana records one of three outcomes for each action policy:

Outcome What it means
dispatched The dispatcher invoked a workflow for the alert episode.
throttled The alert episode matched an action policy but was rate-limited by the frequency setting, so no workflow ran. This is expected behavior, not an error.
unmatched No action policy matched the alert episode. No workflow ran.

unmatched is recorded in the event log but isn't available as an outcome filter in the execution history. To find those records, open Discover and query .kibana-event-log-* with event.provider: "alerting_v2" and event.action: "unmatched".

Note

Episodes that are acknowledged, snoozed, marked inactive, or covered by a maintenance window are excluded before the dispatcher runs and don't appear in the execution history.

The three outcomes above (dispatched, throttled, unmatched) are the event-log terms written to .kibana-event-log-*. The .alert-actions data stream records the same events using different terms. The mapping is:

Event-log outcome (event.action) .alert-actions action_type Meaning
dispatched notified Policy matched, frequency cleared, workflow invoked.
throttled suppress Policy matched but frequency limit not yet cleared; no workflow invoked.
unmatched unmatched No action policy matched the episode; no workflow invoked.

.alert-actions also records triage actions (ack, unack, assign, tag, snooze, unsnooze, activate, deactivate, resolve, unresolve) and the fire action type, which marks that an episode opened or continued. These have no event-log counterpart in this context — they aren't dispatcher outcomes. For the full field reference, refer to Action type values.

Note

suppress in .alert-actions means the same thing as throttled in the event log: the action policy matched but the frequency setting hadn't cleared yet, so no notification was sent. It's unrelated to the eligibility gate that excludes acknowledged, snoozed, or maintenance-window episodes before the dispatcher runs.