DOP-C02 Practice Question 92: AppConfig Feature Flag Rollback
AWS Certified DevOps Engineer – Professional (DOP-C02) · Test 10 · question 2 of 10
A team ships a new checkout feature behind a flag that is read by a Lambda function on every invocation. They want to enable the flag for 10% of traffic, then 25%, 50%, and 100% over the course of an hour, and they want the rollout to stop and revert automatically, without a human watching a dashboard, if the function's error rate spikes partway through.
Which combination of configuration meets this requirement with the least operational effort?
Show answer and explanation
Answer:
AWS AppConfig is built for exactly this pattern: a feature flag configuration is deployed with a deployment strategy that defines the percentage steps and the time between them, and one or more CloudWatch alarms can be attached to the deployment. During the bake period after each step, AppConfig watches the alarms, and if any of them enters the ALARM state, AppConfig automatically halts the rollout and rolls the configuration back to the last deployed version, with no manual intervention.
Option A requires hand-rolled percentage logic in application code and a human to notice the alarm and delete the parameter, which is exactly the operational effort the requirement rules out; Parameter Store has no built-in gradual rollout or alarm-triggered rollback.
Option C adds a DynamoDB Streams fan-out and Contributor Insights paging, both of which still end in a human being paged to intervene rather than an automatic rollback, and neither service provides a percentage-based deployment strategy.
Option D relies on scripted CLI steps and a manual redeploy triggered by a person reading logs; there is no automatic, alarm-driven halt built into a Lambda environment variable update.