Do you mean the best practice in the website Architecture or in it's Disaster Recovery(DR) Architecture? In case you need anything on website architecture then we always have 2 database clusters one to support CM and other to support CD servers. We use clusters here for the high availability.
Now in case of DR, we don't need high availability as it will be used only in case there is a major failure in Main infrastructure and the team will try to get the main infrastructure back on track ASAP. So in DR architecture we mostly have one DB server where the data is cloned or pushed through both the clusters in the Main architecture.
Note: When in DR state, we ask Content Editors to freeze the content as we never clone or push the data vice versa (DR to Main).
Please let me know if this was helpful.