Create an Automation from an Optimization Recommendation
Start with a notification workflow so you can verify recommendation matching before adding remediation.
Prerequisites
- A connected account producing recommendations.
- Permission to create and run workflows.
- A configured notification destination for the test.
Configure the trigger
Create a workflow and add an Optimization Trigger. Select a Cluster, Category, and Rule Name as needed. An empty configuration matches every recommendation; use a narrow filter for your first workflow. See the Optimization Trigger reference for supported values.
The structured trigger does not expose minimum-savings or environment-tag fields. If you need additional logic, evaluate the available payload fields in the workflow before performing an action.
Use the actual recommendation payload
Recommendations arrive under Inputs.event in downstream task parameters:
| Expression | Value |
|---|---|
{{ Inputs.event.recommendation_id }} | Recommendation ID |
{{ Inputs.event.resource_id }} | Resource identifier |
{{ Inputs.event.account_id }} | Cloud account ID |
{{ Inputs.event.category }} | Recommendation category |
{{ Inputs.event.rule_name }} | Recommendation rule |
{{ Inputs.event.estimated_savings }} | Savings value supplied by the recommendation |
{{ Inputs.event.cluster }} | Cluster value |
The payload also contains event_type, cloud_account_id, tenant_id, severity, and status. It does not include an optimization object or proposed CPU/memory request fields. Retrieve and inspect the recommendation details before configuring a rightsizing action; do not treat the resource ID as a Kubernetes workload name.
Add a notification and verify
- Add a notification task from Notification Tasks and select a test destination.
- Include the recommendation ID, resource ID, category, and savings value in the message.
- Test with a representative payload using the validation guide. Dry Run can send real notifications.
- Publish an Active version and make it live.
- When a new matching recommendation arrives, inspect the execution inputs and confirm the notification contains the expected identifiers.
Add remediation after validating the inputs
Choose a task appropriate to the recommendation and explicitly supply its required account, resource, and action parameters. Insert an Approval task where review is needed, and verify decision handling before the change task.
Use Auto Optimize for supported scheduled optimization configurations. Before direct changes, record the previous resource settings and the recovery procedure. Workflow history does not provide a universal one-click infrastructure rollback.