Exchange 2010 - CAS/HT/MBX with high availability

What do you think about what these guys said:

"Casper Pieterse:
 
There is another solution, although, but let me say from the get go, that it isn't supported by Microsoft (yet??) and only works in certain environments. (i.e. low user counts (3000 max)), but it does work.

Instead of using a Windows NLB, create a new cluster resource that contains an IP and DNS name and have your CAS connections connect to that IP/DNS. you can then move the resource to whatever server you think should handle the mapi connections.

I know that the load won't load balance between the two boxes and that the resource is not cluster aware, but it would give you some level of HA without having to buy a 3rd party product or splitting the roles."

and this:

"jader3rd:
 
Casper, that's an interesting idea. I'd never thought of it. It made me think of another possible solution. Just like with Caspers suggestion you wouldn't get load balancing, but you would get HA. I have no idea if this will work, but at the moment I can't think of why it wouldn't. Create a ClientAccessArray and give it the same name as the Dag. This way the IP Address for the NLB will be wherever the cluster group is hosted. You don't need to create another cluster resources, but you don't get load balancing either.
I'd have to test it out to see if there's some naming conflict somewhere, but I could see it working.
Or if there is a naming conflict, create an DNS entry, use that for the Cas Array, but have it resolve to the IP Address of the cluster group. "

I found it on this link: http://social.technet.microsoft.com/Forums/en-US/exchange2010/thread/9826277f-14f4-4e6f-8565-138c6f1eec4b/

What these guys said, does it work? If it does? How do I set this?

Thanks.

Hi,

About Casper idea:

In my opinion this will not work in case of a crash. Outlook MAPI clients do not connect to an IP address but to the CAS server that is associated with the desired mailbox. To connect to this CAS server the clietn WILL use the CAS name. MAPI protocol is able to automatically locate the mailbox and connect to the CAS mailbox associated with it. What will happen if you create a new IP address in your cluster and use this IP to connect to mailbox is (let's suppose your exchange servers are named EXCH1 and EXCH2):
1) Your MAPI client will reach the IP address to open a MAPI session. this will work because there is a CAS service behind this IP.
2) The MAPI client will send an NSPI request to locate the mailbox. The CAS server will answer with the name of the CAS server that is associated with the mailbox (let's say EXCH1).
3) The MAPI client will reopen a new MAPI session to the CAS server EXCH1 to reach the mailbox.

As you can see, if EXCH1 fails for any reason, points 1) and 2) will work but the EXCH1 CAS server associated with the maibox can not be reach at step 3) and access to the mailbox will fail.

Added to that, in case of a Exchange database corruption or service failure (server still running but exchaneg service fail) the new IP cluster ressource would not failover to the other node. So anyway this solution only applies to full server crashes.


About jaer3rd solution:

Again, this only applies for full server crashes, the IP cluster ressource won't failover if Exchange services fail, so you can one day be in a situation where the IP cluster ressource leads you to an Exchange server that is not fully running.

A cluster for Exchange 2010 on Windows 2008 doesn't work the same way as Windows 2003 cluster with Exchange 2003... The Windows cluster service is no more responsible for database failover... Exchange store service IS responsible for database failover in a DAG. So the Windows cluster service can no more provocate failover of other ressource if an exchange database fails...


Have a good day.