The situation--I'm trying to centralize our DHCP services to a single Windows 2008 R2 Server (we're switching to L3 switches that don't provide DHCP services). We currently have Cisco routers at several different sites serving individual subnets for those sites (and acting as DHCP servers), and my goal is to forward a DHCP request at the site to the DHCP server running on a VM in our centralized datacenter.
This should be easy, right? Just setup a matching scope for the subnet on the DHCP server, and add an ip-helper address with the IP of the DHCP server on the interface that will be receiving the DHCP broadcast at the site. Everything I've read and seen indicates that it should be this simple.
Unfortunately, it appears that the forwarded DHCP request packets never reach even the next hop. As part of my desperate troubleshooting, I've setup the router that is the next hop as a DHCP server with matching scopes and ran debugs on both routers (both are Cisco). I can see that a UDP packet from the site router is forwarded from port 67 to port 67 on the next hop router acting as a DHCP server. The next hop router that is acting as a DHCP server, however, NEVER sees the packet come in to port 67. Running a UDP packet tool, I can send packets to port 67 on the router acting as a DHCP server and see them show up there, and packets that are broadcast at the site by Windows clients on port 138 are forwarded to the router and I see them as well.
I've looked at all the router configs and nothing looks like it's blocking anything as far as I can tell. The only other thing I can even think of is to talk to Charter--they provide the fiber service between sites through their routers--to see if they're dropping packets but since other packets are working I'm thinking it might not be them.
Any ideas? Things to check? Something simple I'm missing?
Edit: As luck would have it, we had a Charter technician onsite yesterday to take a look at their equipment that they had an issue with. I asked him if they had anything in their config that would be dropping traffic, he verified that they were simply tunneling traffic from site to site, not touching it at all--it's invisible to us. So that possibility is marked off. I'm thinking that as we switch to L3 switches for the edge at each site, we keep the router to provide DHCP services for the site until we get our core switched over from a router to a L3 switch, I think that may solve our issue.