In HA Kubernetes cluster with stacked etcd topology, wether the data in different etcd member are the same or not?

Question

We use HA Kubernetes by Stacked etcd topology .
It says that:

Each control plane node creates a local etcd member and this etcd member communicates only with the kube-apiserver of this node.

1.1. So the etcd member in Stacked etcd topology will not sync data to each other?
1.2. If so, when a request arrives at LB, it will be forward to a random control plane node. If there are no data on this control plane node, how it works?

2.1. If the etcd member in Stacked etcd topology will sync data to each other, how it works and what's the difference between Stacked etcd topology and External etcd topology ?

Answer 1

I have updated the previously mentioned repository with logs from a test that failed: https://github.com/jfgauron/debug-kubernetes-etcd/tree/main/logs . During that test, the initializing node also ended up dying... The issue is probably related but I'm not sure if it always happens after a certain time or whether one is the cause of the other.

I previously said the failure rate was around 50% but after further testing I'd say it's closer to 20%. I also added a test script to run tests in a loop until it fails.