Alert policy types
This feature is only available in Dagster+.
| Policy type | Triggers on | Notes | Availability |
|---|---|---|---|
| Asset | Asset health - Degraded, warning, or healthy. Sends a notification when the overall health status of the asset changes. Asset materializations - Failure or success. If you use a RetryPolicy, an alert will only be sent after all retries complete. Asset checks - Error, warn, passed, or failure to execute. By default, failed asset checks have a severity of WARN.Asset freshness - Failure, warning, or passing. Sends a notification when the freshness status of an asset changes. Asset metrics - Triggers when an Insights metric aggregated across a selection of assets crosses a configured threshold. See the "Asset metric alerts" note below for additional details. | Asset alerts can be scoped to all assets, a single asset key, or any asset selection. Asset check alerts are sent for any checks on those assets. On Dagster+ Pro, asset alerts also allow you to send alerts to asset owners. | Serverless and Hybrid deployments |
| Run | Triggers on job run success, failure, or time limit exceeded; may optionally include a set of configured tags. | If an alert policy has no configured tags, all jobs will be eligible for that alert. Otherwise, only jobs that contain all the tags for a given alert policy are eligible for that alert. Each newly created organization starts with a long-running run alert policy, which sends an email to the email address used to create the organization when a job run exceeds 24 hours. | Serverless and Hybrid deployments |
| Code location | Triggers when a code location fails to load due to an error. | Serverless and Hybrid deployments | |
| Automation | Triggers when a tick failure occurs for schedules or sensors in the deployment. | Alerts are sent only when the schedule/sensor changes from success to failure, so subsequent failures won't trigger new alerts. You can also configure alerts to only fire after multiple consecutive failures. | Serverless and Hybrid deployments |
| Agent downtime | Triggers when a Hybrid agent hasn't heartbeated within the last five minutes. | Hybrid deployments only. | |
| Insights metric (Experimental) | Triggers when a Dagster+ Insights metric exceeds or falls below a specified threshold over a specified time window. This can be used to alert on:
| Alerts can be scoped to the sum of any metric across an entire deployment, or for a specific job, asset group, or asset key. Alerts are sent only when the threshold is first crossed, and will not be sent again until the value returns to expected levels. Insights data may become available up to 24 hours after run completion. | Serverless and Hybrid deployments |
| Credit budget (Experimental) | Triggers when your organization has reached the monthly credit limit. | Each newly created organization starts with an alert policy of this type, directed at the email address used to create the organization. | Serverless deployments only |
To make health alert notifications as relevant as possible, they only notify once when the health status of the asset changes, instead of every time an event occurs.
For example, assume you have set up an asset health alert to notify when an asset becomes degraded. If the asset is healthy, but fails to materialize after a run begins, then the asset's health status is degraded, and you will receive a health alert notification. If another run for the asset begins, and the asset again fails to materialize, you will not receive a new notification. (An alert configured to notify on materialization failures would continue to send notifications in this case.)
Also note that external assets do not trigger asset health alerts.
Asset metric alerts notify once per alert policy when the threshold is crossed, and will not trigger again until the query conditions are no longer being met. While an alert is "open", the policy will not send additional notifications, even if the metric continues to be out of range or if another asset in the selection later crosses the threshold.
The alert resolves only once the aggregated metric is back within the threshold for the whole selection. When using the ANY asset aggregation, that means every monitored asset must return within range before further notifications will be sent.
To receive a notification for each asset, create a policy per asset scoped to a single asset key instead.