Plan Ingestion and Data Flow Changes
Before changing an existing ingestion flow, identify the metrics, fields, and tags used by queries and alerts. Then choose how to modify the data flow so that the scope of the change matches what you need to verify.
Identify Data Definitions That Must Be Preserved
Before adjusting the processing logic, check which metric names, fields, and tags are used by downstream queries and alerts, and decide which definitions must remain unchanged.
| Situation | Recommendation | Rationale and tradeoff |
|---|---|---|
| The integration provides a default data flow that meets your needs. | Check the default flow and collected data before deciding whether to change it. | Use the existing flow as a baseline to determine whether additional processing is necessary. Avoid changing a working configuration solely for consistency of form. |
| Queries or alert templates already use this integration. | List their metric names and key tags before making changes, and include them in verification. | Renaming or losing related tags can break downstream matching conditions. If a change is necessary, also check the affected queries and rules. |
For example, the EC2 practice in the AWS CloudWatch metrics integration guide requires you to retain metric names and key tags such as region, instance_id, and aws_namespace when using EC2 alert templates. This requirement is specific to that integration; do not assume that all integrations use the same tags. See Out-of-the-Box EC2 Integration for details.
Choose How to Change the Flow and What to Verify
Once you have identified the data definitions to preserve, decide how to adjust the flow and which results will confirm that the published change behaves as expected.
| Situation | Recommendation | Rationale and tradeoff |
|---|---|---|
| You need to retain the original logic for comparison. | Copy the existing data flow before making changes, and identify the version you intend to publish. | A copy preserves the original configuration for comparison but adds another version to track and manage. |
| You need the updated logic to process subsequent data. | Publish the target version and check its deployment status. | Only one version can be deployed for a task at a time. Verify which version is actually running. |
| You have published the change. | Check the target objects, key fields or metrics, and the queries and alert configurations that depend on them. | Deployment status and actual data results are separate checks; verify both. |
For instructions on creating, copying, editing, and publishing data flows, see Task Management. For initial ingestion, see Create a Data Task with a Custom Integration.