# Change one safe resource outside the deployment pipeline

A controlled drift exercise should prove detection, investigation and reconciliation. Stopping at an alert leaves the most consequential part untested.

By Cobnex editorial. Published 2026-09-10. Updated 2026-09-11.

## Choose a reversible test change

Use a non-production resource and a property whose behaviour is understood. A harmless metadata change can test the basic route, while a separate safe configuration change can exercise operational review.

Record the original value and the intended test outcome. Avoid a change that exposes data or disrupts shared services merely to make the alert more realistic.

Apply it through the external path the test is meant to represent, such as a controlled console edit.

## Check what detection sees

Run the normal drift process and confirm that the changed property is within its supported coverage. Inspect the report and the owner notification.

If the tool does not detect the change, determine whether the resource is unmanaged, the property is unsupported or the process failed. Do not treat every missing alert as the same defect.

Check that sensitive values are absent from broad notifications and that authorised reviewers can still access enough detail to investigate.

## Test both possible decisions

First rehearse adopting an intentional live change into the maintained definition. Review the resulting plan and confirm that the next deployment preserves it.

Then use another safe change to rehearse restoring the declared state. Verify the resource's behaviour after correction.

These are different paths. Updating recorded state without changing configuration may leave the next normal plan still proposing a reversal, depending on the tool and workflow.

## Include a changed environment between review and apply

Make a second controlled change after the initial plan is produced. Confirm that the release process detects stale assumptions or produces a fresh reviewed plan as required.

The exercise should not allow an old approval to cover an unrelated new action silently.

Record the complete sequence, including who decided the intended state. The test passes when the difference is detected, understood and reconciled safely through the maintained process. An alert appearing in a dashboard is only the first milestone.

## Sources

- [HashiCorp: managing resource drift](https://developer.hashicorp.com/terraform/tutorials/state/resource-drift)