Exchange 2010 High Availability Setup - Best Practices?

[guest]

We have a single Exchange 2010 SP1 server running mailbox, client, and hub roles.

What are the best practices (or a step-by-step guide) for setting up a second server to host all roles in case the primary server goes down?

I believe I understand DAG well enough but that seems to only take care of the databases - am I missing something? What about the other roles?

Any help here would be appreciated.

[guest]

I wanted to clarify something on the hub role. The hubs do not "take care of themselves." If you have one hub per site, and a hub is down, that site will not get mail delivered to the mailbox server.



so, if you are using a multi-site scenario for DR and resiliency, you will need at each site:



Mailbox server

CAS (more then one)

HUBs (more than one)



The "problem" with the CAS, is that if you have them on the same server as the mailbox server, you cannot place them into an NLB. If they are different AD sites, you cannot place them into an NLB. Therefore, you will need a third party load balancer. Then you need to decide where to put that.. because if it's down then your resiliency is also down =( Ideally you would want something that worked internally and externally. I'm still solving this problem.

With the HUBs.. same situation. a HUB will only deliver mail the to mailboxes in the same site. so, if you have the hub and mailbox role seperate, and your hub is down, you will NOT receive any mail to that site. You would need two hubs, or more. Everytime you add a hub role, you add a whole new exchange license. Which, really sucks.

Lastly, as far as the mailbox servers go, they use failover clustering. You'll have to put your witness at a completely DIFFERENT site in order to only get away with two mailbox servers. Otherwise, the nodes will vote. so your office is down with the witness and one mailbox server the remote site will not automatically kick over because it knows it has lost the vote.

You will need at least three separate locations in order to have a DAG set up for true resiliency. The witness would be placed into the third location. Otherwise, you can put a third mailbox server into the DAG, and then you will have a better situation with node majority. But again, now you are into another exchange license AND an enterprise license for the OS.



They have improved resiliency, but they have far from solved it.

This is a psuedo-rant as we are trying to make a resilient environment, and keep stumbling into these things that are just not very clear in any one location.