We extended the functionallity for the ECS deployments with concourse. After the service gets deployed Concourse would just exit because Terraform doesn’t take the deployment itself into account. This resulted in having false deploys sometimes without any clear reason what happened.
We implemented an extra task at the end of the deploy process that uses the wait services-stable AWS call to check the status of the service after the deployment. When the service doesn’t come up healthy after the deploy this will return a signal that will fail the concourse pipeline and notify in Slack that something went wrong during the deploy.
This functionallity will be rolled out on all ECS services that use Concourse in the next days (A GitHub issue was created in your board to track process of this).