Over the past few weeks I have been busy deploying Exchange 2010. Many of these deployments involved deploying Exchange 2010 on VMWare ESX environments. In a previous post I mentioned how Vmotion and Exchange 2010 DAGs was not supported. However, based on some testing I have changed my opinion of this approach. Perhaps it is just a fluke and I would be interested to hear what everyone out there is experiencing but here goes.
When leveraging DAG on Exchange 2010 and the need to VMotion the server comes up there seems to be an issue with doing a migrate when the Exchange 2010 mailbox server is powered off for the Vmotion. I have seen this occur at two separate client sites. Power off the Mailbox Server, Migrate, DAG broken. Got it?
Now, what I am seeing that if I leave the Passive node online and replicating the passive databases with the active node a live migrate seems to work just fine! I have tested this twice now (which is why I don’t know if it is a fluke or not) and both times migrating the passive node while powered on and replicating caused NO problems. I even went as far as to reboot the passive machine anticipating it to break, but nope, nothing!
So, I’m going to challenge the community out there to try this if you can. Live migrate your passive VM Mailbox node and see what happens. If you can, power it off and migrate and see what happens. I seem to have the same occurrence when the VM is powered down and/or when the VM is powered on.
Either way, from my testing it appears that if you migrate while the DAG member is online (keep in mind I didn’t have any active databases running on the node) it seems to successfully migrate without any problems!
Let me know what your findings are, but this is great news!
**Update** Microsoft has changed their Stance on vMotion and Live Migrate. Both features are now supported on Exchange 2010 SP1. I happen to find this out while at TechEd and did post it here: http://www.scottfeltmann.com/index.php/2011/05/18/microsoft-teched-2011-update-day2/
“Power off the Mailbox Server, Migrate, DAG broken”
And what shows at the time Failover Cluster Management? Is cluster alive?
I’ve seen strange DAG “failures” when witness share was dead.
Since the Passive node is powered down the Cluster is still running but Node 2 is not available. For some reason it only breaks when the passive node is off line and then moved. I have seen this happen two times and two seperate clients!
If you leave the nodes up and running and perform the migration everything works, things don’t break.
When it is broken the cluster shows active but replication is broke.
Did anything change when you powered the system back on? ie. Did the mac address change?
Any other information on why this failed?
Nope, nothing changed. it was very strange
vMotion is now supported
Microsoft source: http://blogs.technet.com/b/exchange/archive/2011/05/16/announcing-enhanced-hardware-virtualization-support-for-exchange-2010.aspx
VMware source: http://kb.vmware.com/kb/1037959