I have a problem whereby my enterprise pool using SNAT and outgoing packets have the new IP of the load balancer. All of this works fine. My SQL Archinging Server/Agent lie on one computer outside of this load balancer pool. When I start front end services, archiving works fine for the first server but the second one does not.
Looking into the outbound queues I see that the Front End servers identify the SQL by the netbios name using DIRECT=OSERVNAME however the SQL Archiving server's outbound shows DIRECT=TCP:[ip of load balancer] So the first connection holds this open and the second one can't connect leaving unacknowledged messages in the queue.
Has anyone had a similar problem? I've disabled NameValidation for OS but that didn't help. So right now I can only use one server in my pool or else it won't archive half the messages.
The OCS front end servers must see their backend services untranslated by NAT ... e.g. MSMQ, SQL, Active Directory. NAT only plays a role between the client->load balancer->front ends. The clients use a LB VIP address and certificate/DNS name, then the LB "sprays" sessions to the FE's.