So I'm sure he question will be asked "why not virtualbox/esxi/xen?", to which I will answer the clients are outlook exchange junkies, so we went with Hyper-V so I could run Trixbox on their windows server. Here's the deal:

Internet -> RB750 (eth1)->Bridge1:192.168.0.1(eth2-5)->eth2 cabled to Netgear 16-port GB switch.
2 ports on switch go to server running Windows 2K8
on server:
eth1:192.168.0.2 - running DHCP
eth2:192.168.0.3 - interface for Hyper-V virtual switch

So when I set this up at first the VM's wouldn't get IPs, but the rb750 was handling the DHCP leases. I figured maybe there's some issue w/other DHCP servers giving IP's to the VM's on the windows server, so I decomed the rb750's DHCP server and configed it on eth1 on the windows 2K8 server. Ta-Da! VM's getting leases, but no route to the internet. All other machines on the LAN are getting the 2K8 leases fine. I can RDP to the VMs from other physical boxes and ping back and forth. However, if I put the eth1(192.168.0.2) of the server back on any port on bridge1, my Vm's lose their leases, but physical machines are fine. So it seems pretty clear that the point at which the packets coming off the virtual switcd are lost is as soon as they hit the rb750

I've tried plugging in the link to eth2 on the 2K8 machine directly into the rb750 and running torch on the appropriate interface, but all I see are winbox mac broadcast packets going to 192.168.0.3. Anything I should try for debugging? Thoughts?

OT, but you probably wont have much luck with asterisk virtualized unless you have an external timing source. Youll have crackly connections.

Never tried with hyper-v, but have with Xen paravirt/fully virt and VMWare. Asterisk requires a hardware timing source which isnt emulated with enough accuracy on virtualization platforms. I ended up having to put a PCI zaptel card in the virtual host and pass it to the guest to get good timing.

Otherwise, I have several Xen Linux/Windows guests behind a RB1000. Works fine. Dont use DHCP, though.

I'm running trixbox (paravirtualized) on my XenServer at home, I have no problems with it. All the servers on the XenServer use DHCP from my RB750 and it is also working fine.

i run trixbox on esx for 2 years now, no issues at all. I have about 20 extensions all over the country, and use it for my primary business lines as well. no issues at all if you have a fast box. no need for a real timer anymore.

Yeah, I had read about clock timing issues on lower CPUs, but this fellah's running dual Xeon E5420 w/16gb of ram, so no problem there.

Any thoughts on how to troubleshoot Hyper-V as opposed to telling me how you all picked the *right* virtualization platform, and are having no troubles?

Maybe Asterisk has it fixed now.

I was unable to do it on a quad core, 2.66 w/ 8gb ram. That was a couple years ago and I havent revisited it. Just put another box on the rack and its been fine since.

Anyways. It was OT.

on the windows box, what subnet masks are you using? You can't have 192.168.0.2 on one interface, and 192.168.0.3 on another if they are both in teh same subnet. Unless you bridge them on Windows as well... Typically for the VM networks you dont assign ANY ip address in windows, only the guests get them.

They are both on /24 subnets. Most configuration guides for Hyper-V give this as the default setup, as the second adapter just get's treated as a virtual switch, but is addressed by the host by it's IP. The 192.168.0.3 adapter does not have a gateway assigned to it, though, to avoid multihoming.

Unfortunately I was unable to resolve this, and as it was part of a client's office install, I had to abandon it. However, I have another RB750 in another enviornment that is passing hyper-v traffic just fine, so I believe it might be some kind of fluke with that particular setup. If anyone else stumbles upon this thread looking for answers, feel free to dm me of email tyler [at] otcpdx (dawt) com