vCloud management cluster is a VMware vSphere High Availability (HA) and vSphere DRS (Distributed Resources Scheduler) cluster that is created to manage a vCloud architecture. A management cluster contains the standard management components, such as ESXi hosts, vCenter Server system, vCloud Director cell servers, database server/s for vCloud Director, and vCenter. A management cluster should have its own shared storage that will store the virtual machines running inside the management cluster. The management cluster should also be separated into a single physical site. We would like to emphasize that for the cloud, it is a must to have a separate management cluster. It is a best practice to place the management components in a management cluster.
You should use vSphere HA and DRS on the management cluster to provide availability for all the management components. For vSphere HA, use the Percentage of Cluster Resources Reserved admission control policy in an n + 1 fashion instead of defining the amount of host failures a cluster can tolerate or specifying the failover hosts. This approach will help you to allow management workloads run evenly across the hosts in the cluster without the need to dedicate a host strictly for host failure situations. But this is not just limited to n + 1; for higher availability, you can add a host for an n + 2 cluster, although doing so is not a requirement of the vCloud private or public service definitions.
You may be wondering why you need a vCenter Server inside your vCloud management cluster. This management vCenter Server will carry clusters that will host cloud workloads. These resources are allocated by vCloud Director as a provider datacenters. Within a distinct vSphere cluster, a provider datacenter translates into a resource pool that is created automatically by vCenter, issued on a request from vCloud Director.
Although you can physically separate the management cluster and resource cluster, it is not a good practice to do so. You should put the management cluster and vCloud consumer resources on the same physical site. If you use a single site, it ensures a consistent level of service. Otherwise, latency issues might arise if workloads must be moved from one site to another.
