The configuration of vCenter HA is pretty simple, and it's done from the vCenter server web client. At this stage, you can't use the new HTML5 interface, and the configuration is done through the older FLEX client:

There are two deployment modes—basic and advanced.

With basic deployment, you need to run vCSA on the same environment that vCSA is managing. If you have a dedicated management cluster where your vCSA is installed, you need to perform advanced configuration since, during the vCenter HA setup, the source vCSA will be cloned, and this won't be possible if the vCSA is not located within the same environment, of course:

Once you have decided what deployment type you will need, you need to set up the HA network. The HA network is used for internal communication between two vCSA servers and the witness appliance.

As the last step, you need to configure settings for the peer vCSA and the witness appliance. On which datastore the appliances will be stored, portgroup association or cluster on which the VMs will be run:

In this case, the configuration wizard complains about using the same datastore as the production VMs, which is not the best practice. You should always use dedicated datastores for your management and the production workloads. Once you hit Finish, you can check your running tasks to see at what stage the deployment is.

Once the deployment is finished, you will see the current state of the cluster, which is the active node, and the state of the passive node and witness appliances:

In the case of hardware failure, the passive appliance becomes active, and you will still be able to manage your environment even if the originating vCSA appliance is no longer available:

Please note that the failover process takes several minutes. If you try to access the web client of the vCSA during the failover, you will see that the failover is in progress.