Question

A company runs a containerized order-processing API on Amazon ECS with the Fargate launch type behind an Application Load Balancer. They use AWS CodeDeploy for blue/green deployments triggered by CodePipeline. The team wants each new (replacement) task set to be validated by an automated integration test suite that sends real HTTP requests to the new version BEFORE any production customer traffic is shifted to it. If the tests fail, no production traffic should ever reach the new version. Which approach meets these requirements?

Accepted Answer

Configure a second (test) listener on the ALB and a CodeDeploy AppSpec hook that runs the integration tests against the test listener; use the AfterAllowTestTraffic lifecycle hook to invoke a Lambda function that runs the suite and returns Failed to abort if tests do not pass.

Answer

Use a canary deployment configuration that shifts 10% of production traffic first, run the tests against that 10%, and roll back automatically if CloudWatch alarms trip.

Answer

Add a CodeBuild test stage in CodePipeline after the CodeDeploy deploy action so integration tests run against the production listener once the deployment completes.

Answer

Configure the BeforeAllowTraffic hook to shift 50% of production traffic to the replacement task set, then run the integration tests, rolling back if they fail.