Our management, an myself actually, wish to test our disaster recovery setup a few times each year. I have a question about our domain controllers, since I would really appreciate not to mess up our AD in the process...
Our setup:
Server room A
Production hosts and production SAN. Here's our "primary" Windows Server 2008 R2 DC VM which holds all the FSMO roles, DNS and DHCP. The catch is that it's also a fileserver and our printserver because when it was created we wanted to save on Windows Server licenses (we still do!).
Because it's an important fileserver then i need to start it in our DR test and that's where I wan't to make sure that I don't mess things up. Since it also has the DHCP server (with a lot of reservations) it is quite an important VM to test.
Server room B
Test/DR servers and SAN's. We replicate the most important VM's to a SAN and all VM's are backup'ed up with our physical Veeam B&R server that is also located in this room.
We have another DC running on test/DR servers so that we still have AD running in case of a disaster in our primary server room.
I have tried this a couple of times, but the first one failed because I couldn't get Veeam B&R working because at that time I didn't have a DC running next to it and it wouldn't work without contact to AD. The next time it worked better, but the replica of our DC didn't want to work without our other DC running so I did that, but hadn't planned things well enough so it looked like I had some issues once I returned to the DC on the production side (this was before Veeam B&R had failback to production).
This is my plan for the next test which should not mess up anything:
1. Shut down all VM's in server room A to simulate complete loss of our production hosts and SAN.
2. Run replication of our primary DC to get all changes into the replica.
3. Use Veeam B&R to failover to test/DR site, starting with the DC and the other important VM's. The DC on the test/DR servers is still running so the "primary" DC should be happy to find another DC to replicate to.
4. Test that everything works.
5. Perform failback to production on all the VM's that have been tested.
6. Commit failback.
7. Start up production VM's again.
By doing it this way I shouldn't get any problems with the two DC's getting our of sync, since all changes that happens during the test is replicated back to production VM.
Is this an OK plan?
I'm planning to use SureBackup to regulary test backups and will also perform manual checks with SureBackup/VirtualLab a few times a year so that we only have to do it full scale maybe once a year.