How to asynchronously scale aurora serverless v2 replicas

The documentation says "When they are idle, they can scale down to the minimum" - have you confirmed in CloudWatch that they really are idle? Handling replication from a busy writer will put some load on a reader but you would hope not too much. Still it's worth looking how busy it is.

I checked and they are not busy as the writer on the heavy workloads. as an experiment I created a provisioned aurora server (with more resources than the serverless ones had), failed over to it and removed one of the serverless instances. and now the serverless instance hasn't gone past 8 acu a single time. if I get the metrics of serverless capacity of the past 24 hours we can clearly see the difference on acu allocation after the changes. And about replication, I might be wrong but my understanding is that with aurora, replication is done at the storage layer and wouldn't put any pressure on reader instances.