Exchange 2010 DAG Automatic Failover Testing/Issue. Not always automatically failing over to healthy DB copy depending on the way mounted server wrecks

Question

Ok I've got 2 exchange 2010 servers that run client access/hub transport/mailbox roles and one exchange 2010 server running just client access/hub transport roles and acts as my bridgehead. The two mailbox servers are running one database setup in a DAG. Server A shows the DB Mounted and Server B shows Healthy. If I reboot Server A via windows GUI Server B switches from healthy to mounted and I see hardly any interruption in service using Outlook 2007. Server A shows "Service down", then "Failed" then "Healthy" and leaves the DB mounted on Server B. This is how it should work, so far so good.

Now if I test Server A being shut down cold, or unplugging both nics from network to simulate failure, Server B switches from Healthy to Mounted and server A switches to "Service Down" but my outlook client never connects to the DB mounted on server B! I can connect to server C (client access/hub transport) and get to my email and even send new email out, but incoming email doesn't deliver until Server A is brought back online and it's DB goes back to Healthy status.

So I don't understand why it auto fail-overs when I reboot the server with the mounted DB copy, causing very little outlook 2007 hiccup if any. But when I shutdown or DC the mounted DB server it DOES mount the healthy copy but outlook 2007 clients can't connect..

I hope the picture I'm trying to paint makes some sense, it's driving me a little batty. Any help would be appreciated!

Solution

I think you need to setup a cas array. Then set the database to use this array. Then do load balancing with a hardware load balancer or 2 dedicated cas servers with nlb. I'll edit this later when at a computer to give more details. This should give you enough to google though.

Comments

[Admin]

Oh and oddly enough if Server B has the mounted copy and server A has the healthy copy, I CAN turn off server B and it fails over to A without issues with outlook 2007 clients connecting...

[Admin]

Is it possible the problem is all three servers are running hub transport/client access roles? The one I'm calling a bridgehead server should be the only one with these roles and just keep server A and B as mailbox servers?

[Admin]

Hey thanks for the response! I sort of found out why it was broken and you're pretty much right on the money. I installed my exchange servers in a reverse order, first installing my mailbox servers with CAS/HT roles included before installing the 3rd HT/CAS server. I thought I could just uninstall the HT/CAS roles from the first two mailbox servers and the exchange environment would be smart enough to configure the organization to use the new and last remaining CS/HT server, but turns out you have to move the responsibility with the management tool before anything works.