Why do you present an an error message to the OWA-user after a database switchover occurs, rather than just change the OWA code to retry the connection to the mailbox server? The current implementation will result in help desk calls during a db switchover...

Robert,
Actually, you might not see any error, depending on the timing of the situation.  The point was that, if the user does get the error, they don't have to log off and log back on.  They clear the error and either refresh, or in some cases do nothing and they are back in business.  In other cases, if they wait long enough, they won't see any error.  In fact, while I was making this final part, several times one of my OWA clients did not see any errors.  I intentionally re-recorded it so folks could see that both browsers can seamlessly re-connect, even if an error appears.  But there were a few times when one of my OWA clients did not produce an error at all and did reconnect automatically in the background without user intervention.  It depends on timing, and user activity.  Hope this helps.

what about the user experience in Outlook 2010, 2007, 2003?

@Exchange: And to giving no error message at all is not possible? Like retrying up to 60 seconds and if reconnect was not succesfull after that, throw the mesage? So that first access after db switchover is just slow, not erroneous.

Pesos, there will be an upcoming blog entry describing the Outlook client experience, as well as the overall client experience various switchover scenarios. Basically, though, the Outlook 2003 and 2007 online experience is one where the user will see the Outlook popup regarding a lost connection and a restored connection once the *over has completed (which happens in 30 seconds or less). The Outlook 2003 and 2007 cached mode experience is to not see this pop-up. In Outlook 2010, you won't see the popup at all. There is a change in Outlook 2010 where it does not notify the user when a connection is re-established in certain cases (for example, in the case of a database switchover to a Mailbox server in the same AD site).

Robert, I'll take that back as feedback. I would not expect any behavior changes before RTM, but I will certainly pass it on to the HA team.

What happens with MAPI connections to the RPC endpoints on the CAS servers if a CAS server is brought down? Does the client automatically find and connect to an RPC endpoint on another CAS box? If so, by what mechanism, and how long does this take? Also, in terms of scalability, we've noticed that mailbox server performance begins to bog down considerably due to the maximum number of simultaneous MAPI connections that a mailbox server can handle. Under Ex2003, we hit this wall with about 2000 mailboxes on a single mailbox server. Under Ex2007, we hit that limit with about 3000 mailboxes. With RPC endpoints located on the CAS server, does this same bottleneck exist there? If so, we would need to deploy considerably more CAS boxes....

Did I hear right? You need Windows 2008 Enterprise for creating DAG?

thx for your answer