The management server itself contains stuff like access control config, the list of clusters, projects, users and who should have access to them, etc.
Individual clusters are self-contained and have the definitions of workloads/services/volumes/etc running in that cluster. (it is not “all in the server DB”, like in 1.x. Conversely, if you have clusters that are running fine and lose the server, the workloads and their definitions are still there).
So you need to back up both.