We have an Exchange 2010 environment configured as follows:
2x Mailbox servers configured in a DAG 2x Hub Transport/CAS servers, configured with a CAS array using NLB
The problem is this - the two servers running the Hub Transport/CAS roles have two NICs, one for general usage and one for the CAS array. We route our external mail via a smart host, which has been configured to accept mail from our specific IP addresses. Our firewall NATs connections from each Exchange Hub server to one of these external addresses. The problem is, traffic (not just SMTP but everything) seems to go randomly out of one NIC, then the other. I can confirm this by going to e.g. whatismyipaddress.com and see it change when I refresh. Both NICs are configured with a default gateway (yeah, I know).
What I have tried so far is removing the gateway from the NIC use for the CAS array. But this then cause client connections (i.e. Outlook) to fail, as our clients are on a range of subnets different to our Exchange servers. I added a static route for our internal subnet range, specifying the CAS NIC as the interface, leaving just the other NIC with a default gateway. But for some reason client connections still fail.
I could configure the smart host to accept mail from the Hub servers on both NICs, but this would mean either assigning additional external IP addresses, or allowing SMTP connections from our generic internet gateway address, which all clients use for browsing. Which I obviously don't want to do.
Is there a way of achieving what I'm trying to do? Other than adding two new servers and separating the Hub and CAS functions? Any ideas why client connections to the CAS array might fail when I remove the default gateway, even though I define a persistent static route? How might I go about troubleshooting this?