Hey all -
NSXv 6.4.x: Consider a 2-node Edge Cluster that is currently setup without vSphere HA. This may have been an oversight by a consulting group who built it or it could be by design. Here are the considerations:
All ESGs/DLRs are deployed in HA so NSX Manager has created DRS anti-affinity rules to keep 0's and 1's away from each other. Good.
All DRS rules are "VMs must not be on the same host". Eh.
In case of an HA event, obviously the NSX ESG/DLR will fail-over within the appliances. However, without vSphere HA the VMs would go disconnected and the ESGs will be SPOF until host restores and someone powers the VMs on. With vSphere HA enabled, the default DRS rule will still result in the ESGs disconnecting/not restarting because HA will not violate the "must" rule.
In VVD it doesn't mention anything for a 2-node edge cluster (perhaps intentionally). I've always worked with a 3-node edge cluster.
Either way my options are basically no vSphere HA. vSphere HA with default "must" rules - aka know effective vSphere HA. Or, reconfigure "must" to "should" but have to do it every time an ESG or DLR is created. I don't know of a way to change the default policy. I guess I could script something, but yeck.
Afaik as of vSphere 6.5, vSphere HA respects affinity rules by default. However, you could always change this behaviour by setting an advanced setting das.respectvmvmantiaffinityrules to false. Then HA will restart the VM, even if that violates the rule.