VCHA is a feature that uses a three-node cluster to protect the vCenter Server from hardware, operating system, or application failures. The three nodes are referred to as active, passive, and witness. VCHA only supports VCSA deployments, not vCenter on Windows, and both embedded and external PSCs are supported. It's important to note that if used with external PSCs, VCHA is not protecting the PSCs—only the vCenter Server itself. Load balanced PSCs would be needed to provide high availability to external PSCs. Keep in mind that it likely doesn't make sense to use vCenter HA if you're not also using load-balanced PSCs, since the idea is to create a highly available management plane.

VCHA is useful when you want to increase vCenter's uptime and you don't necessarily want to only rely on vSphere HA to protect against a host failure. VCHA also protects against service failures. When using an embedded PSC, VCHA will not only monitor vCenter services for failures, but for PSC service failures as well. If a service fails, the passive node becomes the active node. 

There are a few simple system requirements for VCHA. Only one vCenter Server license is required, and the vCenter deployment type must not be tiny due to the additional resources that are required.

There are two options to configure VCHA: Automatic and Manual. If you choose an Automatic configuration, the configuration wizard will clone the existing vCenter Server and configure the cluster networking. If you choose the Manual method, you must clone the vCenter Server and configure the cluster networking yourself. As of vCenter 6.7 Update 1, however, only the Automatic method is available.

After several years of not having a high availability option for vCenter (after VMware deprecated the vCenter Heartbeat product), VCHA provides an internally developed option for those environments that require maximum uptime for their vCenter Server.