#blue_green_deploy

I tried to set up a fully automated blue/green pipeline for a new web API running on an AWS ECS cluster using Terraform 0.15 and the AWS CDK modules. Here’s what I did so far:

• Define two ECS services with the same task definition family but different environment variables.
• Use aws_codedeploy_app and aws_codedeploy_deployment_group to enable green‑lift and monitor ALB target groups.
• Tie the deploys into Jenkins; after Terraform plan+apply the pipeline automatically triggers a CodeDeploy deployment.

Now the problem:

1. During a few recent test releases the target group of the new blue version receives 503s before health checks mark it healthy.
2. Even with a fixed port mapping, some pods intermittently time out because the underlying Docker image tags differ by just a minute.
3. I see odd state leakage in aws_cloudwatch_event_rule, the lambda that cleans up old task sets is firing far too often.

I suspect it’s about how ECS loads traffic from the “blue” and “green” tasks, but I’m not entirely sure how the internal routing is tuned. Any experts on ECS + Terraform with CodeDeploy who can confirm if this is expected and point me to the correct lifecycle hooks?

Would appreciate pull requests if you found a simpler config. Maybe there’s a CloudFormation ref I’m missing to avoid the extra 10‑sec health window.