Scaling ECS Tasks is currently manually managed by the user when they need compute resources. In order to save money from forgotten compute resources, this PR introduces a scale down behaviour that after 20 minutes of no request activity to a given service, the service will be scaled down by -1 Task per 20 minute period.
Context
Scaling ECS Tasks is currently manually managed by the user when they need compute resources. In order to save money from forgotten compute resources, this PR introduces a scale down behaviour that after 20 minutes of no request activity to a given service, the service will be scaled down by
-1 Task
per 20 minute period.cc @rbavery