I'm trying to setup Direct Server Return for my Exchange on-prem front ended by a GSLB setup. I've reviewed countless articles regarding setting up the loopback, metric and weak send/receive.
The problem appears to be that the failover cluster backing the Exchange DAG is including the loopback IP in the cluster network. This causes obvious problems as any heartbeat sent to the loopback would just terminate at the load balancer with no place to go.
Example:
Site A
VIP 10.1.0.10
Exchange ServerA: 10.1.0.11
Loopback ip: 10.1.0.10
Site B
VIP 10.2.0.10
Exchange ServerB: 10.2.0.11
Loopback ip: 10.2.0.10
Cluster Networking:
Cluster Network1:
ExchangeA Ethernet0 (10.1.0.11)
ExchangeA Loopback (10.1.0.10)
Cluster Network2:
ExchangeB Ethernet0 (10.2.0.11)
ExchangeB Loopback (10.2.0.10)
I'm able to see with captures that heartbeat requests are sent to 10.1.0.10 and 10.2.0.10. They, of course, arrive at the load balancer and are dropped as they aren't valid https,smtps traffic.
I'm at a loss why others haven't had this issue (I see a few similar unresolved articles around the web). The only thing I can think of to resolve the issue is to keep the loopback and it's "spoofed" ip out of the cluster networks but failover clustering doesn't seem to support that.
