What is environment drift?
Environment drift is the accumulation of small changes in applications and configuration that eventually stops an automation working. No single change is significant, and none of them was made with the automation in mind. The automation continues until the aggregate crosses whatever tolerance it had.
Also called: Configuration drift
Why nothing changed and the automation stopped anyway
The incident report says that nothing changed, and it is very nearly true. A browser updated itself overnight, and a field label was retitled in a quarterly release. Each change passed its own approval, and neither was assessed against a dependency nobody had recorded. Drift is why an automation estate needs continuous verification rather than a test at acceptance.
- Automation dependencies are rarely recorded, so change processes cannot assess them.
- Each contributing change is individually too small to attract review.
- The failure appears at a point unrelated to the change that caused it.
- Acceptance testing proves a state that begins expiring immediately.
How environment drift works
- Every automation declares the applications and the versions it depends upon.
- Those dependencies are registered so change processes can consult them.
- Synthetic transactions exercise each critical path on a schedule against production.
- Environments are described as configuration and compared against the declared state.
- Deviations are reported before the next scheduled execution rather than after it.
- A repair updates the declared dependency alongside the automation itself.
How environment drift is evaluated
Drift is measured by how quickly a divergence is detected rather than by how often it occurs. Examine the following across the estate:
- time between an environment change and detection of its effect
- proportion of automations with declared dependencies
- frequency of synthetic transactions against each critical path
- deviation between the declared and the deployed environment state
- failures first detected by a business user rather than by monitoring
Limitations, risks and common failure modes
- Dependency registers are completed at delivery and never updated afterwards.
- Synthetic transactions exercise the ordinary path and miss the branch that broke.
- Automatic browser updates fall outside the change process entirely.
- A certificate renewal fails only the automation, so nobody connects the two events.