Question

A team uses CodePipeline with a CodeDeploy blue/green deployment to Amazon ECS on Fargate. The service runs behind an Application Load Balancer with a production listener and a separate test listener. Currently the deployment shifts 100% of traffic to the replacement (green) task set immediately after it passes health checks. The team wants to reduce risk by shifting only 10% of production traffic to the green task set, waiting 5 minutes, and then shifting the remaining 90%, with automatic rollback if a CloudWatch alarm triggers during the shift. Which change best achieves this with the least custom code?

Accepted Answer

Change the CodeDeploy deployment group to use the predefined CodeDeployDefault.ECSCanary10Percent5Minutes deployment configuration and associate the CloudWatch alarm in the deployment group's alarm configuration.

Answer

Add a Lambda function invoked by the AfterAllowTestTraffic hook that manually calls the ELB ModifyListener API to weight the target groups at 10%, then 90% after sleeping 300 seconds.

Answer

Switch the deployment group to use CodeDeployDefault.ECSAllAtOnce but enable the deployment group's automatic rollback on alarm so traffic shifts gradually.

Answer

Configure an ECS rolling update with a minimumHealthyPercent of 10 and a deployment circuit breaker to gradually replace tasks over 5 minutes.