silinternational
Make deploys work when > 100 containers
We ran into a problem over at @sparkpost with AWS ECS where AWS' API was returning nextPage tokens that were invalid, and this affected us downstream with ecs-deploy because the list-tasks command failed on a service we run with over 100 container instances. But looking more into the code, I realized that the later call:
$AWS_ECS describe-tasks --cluster "$CLUSTER" --tasks $RUNNING_TASKS
would fail with this service because that --tasks command can take 100 container IDs max. So ecs-deploy seemingly cannot work for ECS services with > 100 container instances.
This PR refactors the logic somewhat substantially, but there's a way to call into the describe-service command and pull the status of deployments, ensure that the new service successfully launched all containers, and then ensure that there is only 1 active deploy left. I think this logic mirrors the original logic, but since it only calls the describe-service API, I think this will be gentler on AWS services as well as having a narrow surface area for errors on AWS' end. And it'll also work for larger fleets of container instances 😀
@fillup I don't know why this problem is still not merged. In order to use ecs-deploy even when operating heavy infrastructure, it must be merged. Please review.
Given the age of this PR and that it has conflicts, I'm going to close it out. If you still want to pursue this, please submit a new PR based on the latest code.
