I have reviewed the documents in Technet describing how to deploy DHCP Failover using Windows DHCP service. There are some details that are missing that I need answered.
Our topology is 6 branch offices and 1 HQ office. These locations are linked together by site to site VPN tunnels in a hub and spoke arrangement. We would put a DHCP server at HQ and in each branch office. Our goal is for each branch office server to serve local clients and rely on the HQ server as a hot standby. We would also want HQ server to use one of the branch office servers as its hot standby. That would create a unique situation where one branch office server (call it Branch Office #1) and the HQ server would both use the other as a hot standby, assuming that is possible with this feature.
I am unclear on how clients in a branch office actually contact the hot standby server if the branch office's dhcp server is down. Typically a client sends a broadcast packet into its subnet looking for a DHCP server. When the local DHCP server is down, how do the packets get through the VPN tunnel to the HQ hot standby server?
Typically the way DHCP requests cross subnets is by using a DHCP relay, typically running on a router or other piece of networking hardware. However, my understanding is also that you should not have a DHCP relay and DHCP server running in the same subnet. It seems like that would be functionally the same as having two DHCP servers in the same subnet. Client requests would be sent to the local DHCP server and relayed to the hot standby by the DHCP relay.
The alternative would be to change the behavior of the router so that dhcp broadcast packets from clients in a branch office would be retransmitted the broadcast address of the HQ server's subnet. This seems like the same problem as before.
Am I missing something here? Is this type of configuration even supported?