Alert system tables reference

Important

This system table is in Public Preview.

This page includes information on the alert system tables, including an outline of each table's schema. Use these tables to query your workspace alerts and their evaluation history directly in SQL, so you can audit alert definitions, analyze evaluation trends, and monitor alert workloads at scale.

The system.alert schema contains two tables:

• system.alert.alerts: The configuration of every alert, including its definition, schedule, evaluation thresholds, subscribers, and lifecycle timestamps.
• system.alert.alert_evaluation_history: One row per alert evaluation, capturing the evaluated state, result values, notification delivery status, and error details.

Requirements

By default, only users with both the account admin and metastore admin roles have access to the alert system tables. To share a table's data with a user or group, Databricks recommends creating a dynamic view for each user or group. See Create a dynamic view.

Alert configuration table schema

The system.alert.alerts table is a slow-changing dimension table. Each row captures the configuration of an alert at a point in time, so a single alert has multiple rows when its configuration changes.

Table path: This system table is located at system.alert.alerts.

The system.alert.alerts table uses the following schema:

Column name	Data type	Description	Example
account_id	string	The ID of the account.	23e22ba4-87b9-4cc2 -9770-d10b894b7118
workspace_id	string	The ID of the workspace where the alert is defined.	1234567890123456
alert_id	string	The ID of the alert.	2762099691916865
display_name	string	The display name of the alert.	Test Alert
owned_by	string	The username of the alert owner. Set to Unavailable if the user has been deleted.	example@databricks.com
compute	struct	A struct that represents the compute resource attached to the alert. The type value is either WAREHOUSE or SERVERLESS_COMPUTE.	{ type: WAREHOUSE, cluster_id: NULL, warehouse_id: 802f6d5283291c0d }
run_as	string	The ID of the user or service principal whose credentials are used to run the alert.	2967555311742259
schedule	struct	A struct that represents the alert's schedule, including its pause status, cron schedule, and time zone.	{ pause_status: UNPAUSED, quartz_cron_schedule: 0 0 10 * * ?, timezone_id: UTC }
evaluation_source	struct	A struct that represents the source column from the query result used to evaluate the alert, including the aggregation applied to it.	{ aggregation: SUM, display: x, name: x }
evaluation_comparison_operator	string	The operator used to compare the source value against the threshold during evaluation. For example: GREATER_THAN, LESS_THAN, and EQUAL.	LESS_THAN
evaluation_threshold	struct	A struct that represents the threshold the source value is compared against. The threshold is either a static value or another column.	{ value: { double_value: 1.25 } }
evaluation_empty_result_state	string	The state the alert reports when the query returns an empty result. Possible values are OK, TRIGGERED, and ERROR.	OK
notify_on_ok	boolean	Whether to notify subscribers when the alert returns to the OK state.	true
retrigger_seconds	int	The number of seconds the alert waits after being triggered before it can trigger again. If 0, the alert does not trigger again.	0
subscriptions	array	An array of the subscribers that receive the alert notification. Each subscriber is identified by a user email or a notification destination ID.	[ { user_email: example@databricks.com, destination_id: null } ]
change_time	timestamp	The time the configuration was last changed. Time zone information is recorded at the end of the value with +00:00 representing UTC.	2023-01-01T01:01:01.123+00:00
create_time	timestamp	The time the alert was created. Time zone information is recorded at the end of the value with +00:00 representing UTC.	2023-01-01T01:01:01.123+00:00
delete_time	timestamp	The time the alert was permanently deleted. Alerts moved to trash are not recorded. Time zone information is recorded at the end of the value with +00:00 representing UTC.	2023-01-01T01:01:01.123+00:00

Alert evaluation history table schema

The system.alert.alert_evaluation_history table records one row for each alert evaluation, capturing the evaluated state and related statistics.

Table path: This system table is located at system.alert.alert_evaluation_history.

The system.alert.alert_evaluation_history table uses the following schema:

Column name	Data type	Description	Example
account_id	string	The ID of the account.	23e22ba4-87b9-4cc2 -9770-d10b894b7118
workspace_id	string	The ID of the workspace where the alert is defined.	1234567890123456
alert_id	string	The ID of the evaluated alert. Joins to alert_id in system.alert.alerts.	2762099691916865
job_id	string	The ID of the job that ran the alert. Joins to job_id in system.lakeflow.jobs. NULL when the alert runs on its own schedule instead of as part of a job. Use a LEFT JOIN when you query against system.lakeflow.jobs.	906851285941474
task_key	string	The reference key for the task within the job. Joins to task_key in system.lakeflow.job_tasks. NULL when the alert runs on its own schedule instead of as part of a job. Use a LEFT JOIN when you query against system.lakeflow.job_tasks.	evaluate_alert
alert_run_id	string	The ID of the alert evaluation run.	4104302562320643
evaluated_state	string	The evaluated state of the alert. Possible values are OK, TRIGGERED, and ERROR.	TRIGGERED
evaluated_source_result	struct	A struct that represents the source value produced by the evaluation.	{ double_value: 1.25 }
evaluated_threshold_result	struct	A struct that represents the threshold value from the evaluation, or the static value set in the alert configuration.	{ double_value: 1.25 }
error_code	string	The error code if the evaluation failed with an error.	INTERNAL_ERROR
error_message	string	The user-visible error message if the evaluation failed with an error.	Query execution failed
notified_subscriptions	array	An array of the subscribers that were successfully notified. Each subscriber is identified by a user email or a notification destination ID.	[ { user_email: example@databricks.com, destination_id: null } ]
failed_subscriptions	array	An array of the subscribers whose notifications failed to send. Each subscriber is identified by a user email or a notification destination ID.	[ { user_email: example@databricks.com, destination_id: null } ]
start_time	timestamp	The time the evaluation started. Time zone information is recorded at the end of the value with +00:00 representing UTC.	2023-01-01T01:01:01.123+00:00
end_time	timestamp	The time the evaluation ended. Time zone information is recorded at the end of the value with +00:00 representing UTC.	2023-01-01T01:01:01.123+00:00

Sample queries

The following sample queries show common ways to analyze alert configuration and evaluation history. Each query uses named parameter markers for the alert and workspace IDs, so you're prompted for the values when you run the query.

Find the most recent evaluation for an alert

This query returns the latest configuration of a specific alert alongside its most recent evaluation.

SELECT
  c.alert_id,
  c.display_name,
  c.compute,
  c.schedule.quartz_cron_schedule AS schedule,
  c.evaluation_comparison_operator AS operator,
  c.evaluation_threshold.value.double_value AS threshold,
  c.change_time AS config_last_updated,
  e.evaluated_state,
  e.start_time AS evaluation_time,
  e.end_time,
  TIMESTAMPDIFF(SECOND, e.start_time, e.end_time) AS eval_duration_seconds
FROM (
  SELECT *
  FROM system.alert.alerts
  WHERE alert_id = :your_alert_id
    AND delete_time IS NULL
  ORDER BY change_time DESC
  LIMIT 1
) c
LEFT JOIN (
  SELECT *
  FROM system.alert.alert_evaluation_history
  WHERE alert_id = :your_alert_id
  ORDER BY start_time DESC
  LIMIT 1
) e ON TRUE;

Find which alerts triggered most in the last seven days

This query counts how many times each alert triggered in a workspace over the last seven days.

SELECT
  alert_id,
  COUNT(*) AS trigger_count_7d
FROM system.alert.alert_evaluation_history
WHERE workspace_id = :your_workspace_id
  AND start_time >= CURRENT_DATE() - INTERVAL 7 DAY
  AND evaluated_state = 'TRIGGERED'
GROUP BY alert_id
ORDER BY trigger_count_7d DESC;

Track how an alert's state changed over time

This query summarizes the daily count of each evaluated state for a specific alert over the last 30 days. Use it to build a dashboard that monitors alert reliability.

SELECT
  DATE_TRUNC('day', start_time) AS eval_day,
  COUNT_IF(evaluated_state = 'OK') AS ok_count,
  COUNT_IF(evaluated_state = 'TRIGGERED') AS triggered_count,
  COUNT_IF(evaluated_state = 'ERROR') AS error_count
FROM system.alert.alert_evaluation_history
WHERE alert_id = :your_alert_id
  AND start_time >= CURRENT_DATE() - INTERVAL 30 DAY
GROUP BY 1
ORDER BY eval_day ASC;

List active alerts in a workspace

This query returns the latest configuration of each alert that is not deleted and not paused in a workspace.

WITH latest AS (
  SELECT
    *,
    ROW_NUMBER() OVER (PARTITION BY alert_id ORDER BY change_time DESC) AS rn
  FROM system.alert.alerts
  WHERE workspace_id = :your_workspace_id
    AND delete_time IS NULL
)
SELECT
  alert_id,
  display_name,
  compute,
  schedule.quartz_cron_schedule AS cron_schedule,
  schedule.pause_status
FROM latest
WHERE rn = 1
  AND schedule.pause_status != 'PAUSED';