Must AWS Neptune reader and writer DB instances in the cluster be of the same size?
In the console I noticed I can modify them to be difference instance types. What will happen if the writer is very powerful, and a lot of data is written to the graph DB, but the reader is the smallest instance type?
I think the reader (read replica) instance will fall behind since it cannot replicate that fast?
It seems that under the hood, Neptune uses similar RDS-type DB instances so maybe it's the same with any RDS replicated DB like Aurora too..
Neptune does support heterogeneous clusters, which basically means that you can have instances of different types in the same cluster. It is a pretty common use case of having a really powerful master and smaller replicas, for various sub applications.
Replication lag across readers is mainly driven by network bandwidth on the instance and the amount of catch-up (data) that needs to be done by the replica. While instance types do have differences when it comes to network bandwidth, usually it is not significant enough to impact replication. If you do see that happening, then its best if you scale up the replica instances to get better network performance. The recommendation is to use replicas which are at least as powerful as the master. Since failover can happen outside of your control, homogeneous clusters end up being the ideal choice for most scenarios.