First of all, what I’m trying to achieve is scale-out/scale-in during deployment with aws update-service so that instance would not require double the memory of service (which is quite demanding memory-wise) at all times.

I do not provide Capacity Provider Strategy in update-service because cluster has a default Capacity Provider Strategy set.

Everything, including ASG, cluster, ECS services, Capacity Provider and Capacity Provider Strategy is Terraformized. 1 ASG desired capacity set in Terraform and max capacity is 2.

My issue is that currently I have 2 instances in ASG (set to 2 desired instances by Capacity Provider), both being vastay underutilized. Yet CapacityProviderReservation metric in the CloudWatch is reported as 135, meaning that it would scale-out if max desired capacity was not 2. And I’d actually expect a scale-in to happen, because all 4 services that are now spread accross 2 instances could fit into 1.

Why is it not scaling-in? Are my expectations on how Capacity Providers work incorrect? Or maybe there are other ways to achieve what I’m trying to achieve?