Hi all. So this might be old news for many, but I just discovered this feature of 2012 R2 and think it is just great. I often wondered why Microsoft required shared storage and the clustering role to grant high availability to the DHCP service. Well, not anymore. Credit to This Site is due. Scott pretty much says it all, but I will key off one point. In his post he mentions the following regarding the forwarding of DHCP requests from a client across subnets using a DHCP relay agent, sometimes called “helper” agent.
“This is all good, but what happens in the case where we have more than one DHCP server, and require multiple DHCP Relay Agent addresses at the router? This is a great question, and it explains why we will really appreciate the new Windows Server 2012 R2 DHCP services.”
He doesn’t really mention how this should be configured in such a case. I expected the two DHCP servers to offer a Virtual IP Address (VIP) like clustering would give you, but unfortunately they do not. So the problem exists when you have two DHCP servers in the load balancing or failover mode, but you are only forwarding DHCP requests to one of the server IP’s. In the case of load balancing I would imagine this would work very poorly as the 2nd server would never get to take the load as expected. I wonder if the primary server would simply not answer requests intermittently? Something to lab on another time.
In the case of a failover configuration, it would still take manual intervention to update the relay agent IP address.. Which really isn’t any better than traditional split scope implementations where the 2nd scope remains in a disabled state. So it really comes down to the routing/switching device to forward DHCP requests to the server that is actively servicing DHCP. Or is “alive” in other words. There are plenty of devices out there that do this, but the device I have used specifically with this feature is the Sophos UTM 9. Mainly because they give a free home use license for the lab. I was able to create a network object that contained each DHCP server IP. The UTM would designate the primary DHCP server in the failover pair as primary and would “ping” each every 60 seconds. Once the primary server no longer responded to ping the UTM would change the IP of the network object to the secondary server. When configuring UTM DHCP relay you would simply use the fialover network object rather than one server or the other.
There have to be other devices out there that can do something like this, SonicWALL, Cisco, HP Procurve… Anyone else know of a comparable product and feature to support DHCP relay with multiple DHCP servers?
This setup is not compatible with load balancing, I can’t quite figure out how this would work in the case of multiple subnets and DHCP relay.
Thanks for reading.