Hello, We have an Exchange 2003 setup in backend frontend topology. We are in the process of transitioning to Exchange 2010. I finished installing CAS and is in the process of installing HT role. During install I have a windows for "Mail flow Settings". It says to enable mail flow to Exchange 2003, a routing group connector is required. Select an Exchange 2003 server in the routing group to which exchange 2010 will connect. Once I hit browse I see my backend server name (Virtual name as it is clustered) and two frontend server names (network load balanced). I am confused as of what to use here. My current Exchange 2003 ESM configuration is as follows: Administrative Group > HQ > Routing Groups > HQ > Connectors > Outbound Connector I will appreacite if you could explain what is expected here and the best practice? Thanks in advanceTom

You actually need two routing group connectors, one for each direction between Exchange 2003 and Exchange 2010. From my experience, Exchange 2010 doesn't like the RGC to be terminated on an Exchange 2003 cluster. Do your front-end servers also serve as bridgeheads? That is, can they route mail? If so, then there's no reason not to connect he RGC to both of them since you can specify more than one for redundancy. Do not try to connect it to the load-balanced address.Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."

Thanks Ed. With our Exchange 2003 setup, Outbound mail connector on Exchange 2003 backend server is set to forward all messages to a third party SMTP server installed in our network. We use Exchange 2003 backend server as bridgehead server with this Outbound mail connector. Also, with Exchange 2010 setup we have two CAS/HT servers. I have a load balancer exclusive for outlook web application (with external access) and a load balancer for CASArray which will serve as MAPI endpoint (only internal access). I read from technet that Microsoft support NLB with Exchange 2010 SP1 thus I was planning to use the cluster name of the load balancer I setup for OWA. I know it is confusing but let me brief it: Exchange 2003 in a frontend backend topology. I have two CAS/HT Servers with each roles installed on two servers. Each server has 4 FOE adapters.I set these as a team of two adapters thus I have one team servicing MAPI/RPC called "MapiArray" as clustername in NLB and other servicing OWA called "OWA" as clustername in NLB. I like to forward all messages to OWA IP for outbound traffic. MAPIArray is our CASArray name (NON-RESOLVABLE PRIVATE ADDRESS) and OWA (RESOLVABLE PUBLIC ADDRESS VIA NAT'g)is a HA NLB cluster. Please advice on how I could get this configuration functional. ThanksTom

Typically the convention is to use the FE because the FE is what's typically designated when creating RGC and is also the routing group master. Launch ESM; Expand Administrative Group;expand Routing Group;Expand First Routing Group;Click on Members;right pane designated master. You also need to supress link state on Exchange 2003. Suppress Link State Updates http://technet.microsoft.com/en-us/library/aa996728(EXCHG.140).aspx Example setup: Cas Array Name: outlook.domain.local (don't put this in your SSL SAN cert, add this DNS record to AD DNS 10.0.0.10) NLB Name: outlook.domain.com (needs to be in your SSL SAN cert, add this DNS record to AD DNS 10.0.0.10) James Chong MCITP | EA | EMA; MCSE | M+, S+ Security+, Project+, ITIL msexchangetips.blogspot.com

Thanks James. In m case, we have three servers listed under Exchsrvr 2003 RGC with Exchange 2003 backend server virtual name as the master and the NETBIOS name of two Exchange 2003 frontend servers as members. During Exchange 2010 HT role setup, I used Exchange 2010 Backend server virtual name in "Mail flow settings" window. Is it a recommended way during the coexistence period of Exchsrvr 2003 and Exchsrvr 2010? Also, does Exchsrvr 2010 OWA NLB virtual name be configured as CASArray? I understand a site can have only one CASArray. My understanding about CASArray is " A group of CAS servers joined together as a common name using a load balance solution to provide HA for internal MAPI/RPC connections". This CASArray is not required for Exchsrvr 2010 OWA to function. Please correct me if i am wrong. This will clear a lot of confusions about the relationship between CASArray and OWA using Microsoft NLB. Thanks in advance.Tom

As far as is it recommended there's not a hard answer because it depends on your own scenario and each orgs requirement. Both will work, some guidelines will choose the backend because you will eventually decom your 2003 FE before the 2003 BE so you don't want to rehome the conector later again to use the BE. Other orgs may still choose to do the FE anyways because they don't want to control mailflow in the same path or for additional load reasons on the BE etc. Below is a guideline its for 2007 but high level is still the same. Exchange 2000/2003 to Exchange 2007 migration guidelines http://blogs.technet.com/b/mbaher/archive/2008/06/04/migrating-exchange-2000-2003-to-exchange-2007-guidelines.aspx Choose the NLB name to be the public NLB name ie mail.domain.com and not your CAS array name. Make sure your CAS array name is not resolvable from the internet. Yes your understanding is correct CAS array is not required but recommended for load balancing and high availability. James Chong MCITP | EA | EMA; MCSE | M+, S+ Security+, Project+, ITIL msexchangetips.blogspot.com

Thanks again James for the clarification. So if I understood this, CASArray is not required for MAPI endpoint or OWA. It is recommended for HA and load balancing. In my case I have four adapters. I want to load balance both MAPI and OWA. Here is my design: Server1 (Four FOE adapters) NIC1 + NIC2 (Teamed) -> NLB_MAPI NIC3 + NIC4 (Teamed) -> NLB_OWA Server2 (Four FOE adapters) NIC1 + NIC2 (Teamed) -> NLB_MAPI NIC3 + NIC4 (Teamed) -> NLB_OWA FQDN for load balanced MAPI connection is MAPIArray. This has a DNS entry in Internal DNS Server and is not resolvable. I also set this common name "MAPIArray.domain.local" as the ClientAccessArray for MAPI endpoints.This name will be used for MAPI clients to connect to Exchange Backend /MBX servers. FQDN for load balanced OWA connection is OWA. This has a DNS entry in internal and external DNS Server and is resolvable from internet. I don't have a ClientAccessArray set for OWA NLB as ClientAccessArray is limited to one persite. Now that OWA is only NLB when Server1 fails, all external clients connected via Internet to Server1 is disconnected from OWA with some sort of HTTP not available error code. If Server2 is available, re-enering the page with credentials will log the user to OWA. Please correct me if my understanding is not correct. Thanks Tom

Your second point "I don't have a ClientAccessArray set for OWA NLB as ClientAccessArray is limited to one persite." There is no such thing as CAS array for OWA. For OWA it's using traditional NLB or Hardware load balancing. Yes if OWA node A fails those users connected will get disconnected and they have to reauthenticate which will take them to node 2. The same goes for Outlook clients since they connect to the CAS array mapiarray.domain.local if node A goes down they will get a password prompt and they have to reauthentic. James Chong MCITP | EA | EMA; MCSE | M+, S+ Security+, Project+, ITIL msexchangetips.blogspot.com