Best practices / On premise architecture


I tried to find some answers in the forums but I did not find answers to my questions.

We are using a managed Kubernetes service since few years.
Now, we must deploy the same service on premise and… we have few experience with Kubernetes on premise.

We have currently 3 datacenters without extended network. We can configure an extended network between two of us with a low latency (less than 5ms). The third datacenter is more far with an higher latency: 20ms. But we can not extend a network to this third datacenter.

We would like to install the worker nodes on the 2 nearest datacenters only.
We would like to install the control plane on the 3 datacenters (2+2+1 master nodes).

What are the inherent risks that we could encounter on the controlplane with this configuration (no extended network on all datacenters) and with the latency?
Have you some best practices to share?