Monitor alerts

This page describes how to create a Databricks SQL alert based on a metric from a monitor metrics table. Some common uses for monitor alerts include:

  • Get notified when a statistic moves out of a certain range. For example, you want to receive a notification when the fraction of missing values exceeds a certain level.

  • Get notified of a change in the data. The drift metrics table stores statistics that track changes in the data distribution.

  • Get notified if data has drifted in comparison to the baseline table. You can set up an alert to investigate the data changes or, for InferenceLog analysis, to indicate that the model should be retrained.

Monitor alerts are created and used the same way as other Databricks SQL alerts. You create a Databricks SQL query on the monitor profile metrics table or drift metrics table. You then create a Databricks SQL alert for this query. You can configure the alert to evaluate the query at a desired frequency, and send a notification if the alert is triggered. By default, email notification is sent. You can also set up a webhook or send notifications to other applications such as Slack or Pagerduty.

You can also quickly create an alert from the monitor dashboard as follows:

  1. On the dashboard, find the chart for which you want to create an alert.

  2. Click Kebab menu in the upper-right corner of the chart and select View query. The SQL editor opens.

  3. In the SQL editor, click Kebab menu above the editor window and select Create alert. The New alert dialog opens in a new tab.

  4. Configure the alert and click Create alert.

Note that if the query uses parameters, then the alert is based on the default values for these parameters. You should confirm that the default values reflect the intent of the alert.

For details, see Databricks SQL alerts.