Since Vcenter is used to configure NSX Manager which is responsible for Management Plane, if Vcenter is down on either site would affect provisioning of local or global Logical switches through GUI, and assigning these port groups to VMs, since NSX Manager needs Vcenter, which is the dVS management plane for programming the distributed port groups on distributed vSwitches.
If Primary Site Vcenter is down, GUI of Primary NSX manager is not available, therefore creation of other Universal objects such as UDLR, UdFW Rules may not be possible. Also global objects within the scope of Primary Site1 may not be possible. Cross-Vcenter vmotion may not be possible, but vmotion within site2 exptected to continue.
The case of Secondary Site Vcenter down, creation of global objects within the scope of Site2 may not be possible. Cross-Vcenter vmotion may not be possible, Site1 vmotion within Site1 expected to continue. In the rare case that Site1 fails during Site2 vcenter down, it may be possible to promote Secondary NSX Manager to primary role with Rest API or VRO Automation:
http://solutionyst.com/vmware-nsx-promoting-a-secondary-nsx-manager-with-the-rest-api-in-cross-vcent...
http://www.routetocloud.com/2016/09/nsx-cross-vc-extensibility-kit/
Trafficwise the VM traffic should not be effected since Vcenter is not on the data path, and NSX Manager, Controller Cluster is functioning. Since vmotion is not possible locally on the site that vCenter is down VMs would stay on ESX hosts that they work on until vCenter is available again. If an ESX host also goes down during this time, HA would still work even Vcenter is not available, and Controllers would update their VTEP and MAC tables which would allow for the VMs recoverd through HA continue to forward traffic.
vCenter Server down - what functionality is lost?
