Question

A team deploys a Lambda function through AWS CodeDeploy using a canary configuration that shifts 10% of traffic for 5 minutes before completing. During a recent deployment, the new version returned a spike of errors, but CodeDeploy still completed the full traffic shift because no automatic rollback occurred. The team wants future deployments to automatically roll back when the new version misbehaves during the canary window. What should they configure to achieve this with the least custom code?

Accepted Answer

Associate a CloudWatch alarm on the Lambda function's error metric with the CodeDeploy deployment group and enable rollback on alarm.

Answer

Change the deployment configuration to LambdaAllAtOnce so failures are detected immediately after cutover.

Answer

Add a PostTraffic lifecycle hook Lambda that queries CloudWatch and manually calls the CodeDeploy StopDeployment API.

Answer

Increase the canary interval to 30 minutes so more errors accumulate before the traffic shift completes.