I haven't seen anyone describe this scenario, so maybe it can't be done, but it seems it should work. Can someone either confirm this is viable or set me straight and tell me it's not wise (and if not, why not).
I currently have SBS 2008 (2008 Server + Exchange Server 2007 + some other components). I use redirected folders so all my user data is stored primarily on the server, and then sync'd with their local harddrives for roaming use and as an additional backup precaution. I want to upgrade to a modern version of Exchange (don't care too much about the underlying server OS, but seems like a good time to upgrade everything). I also want to improve the reliability for rapid recovery in the event of a system failure. I have always felt nervous with SBS because it only runs on a single server.
My plan is:
- Install Windows Server 2012 Standard on a new server machine
- Join that new server to the existing SBS domain and sync the users, etc.
- Install Exchange Server 2013 on the new 2012 Server
- Migrate the existing Exchange data and account info to the new server
- Demote the SBS Server and promote the new server to primary DC
- Reformat the old SBS server and install Windows Server 2012 on it too, and join it to the domain as a second DC
- Set up DFS Replication and keep all my user data folders between the two servers (so if one server fails, the other can take over, just not automatically like with a cluster)
- Install Exchange Server 2013 on the reformatted 2012 Server box (so second instance on the domain) and set it up to sync with the first Exchange Server 2013 via DAG (so if either computer crashes, the other can take over for both Exchange Server and file services)
Wouldn't that give me a complete slow-response failover system without the need for setting up a cluster (I know a cluster would provide instant failover protection, but I don't need that kind of speed - a manual switchover in the rare event of a server failure is acceptable for us)? I could do this using cheap hardware for each server, because unless both servers fail at the same time, clients can point to the other server. And because it's not a formal cluster, I don't need to buy anything for a shared SAN. This seems like such an obvious more cost-effective solution for a small network, I don't understand why this isn't a common approach, which makes me wonder if there is some reason it won't work.
Any suggestions or feedback? Are Redirected Folders still recommended for users on a Server 2012-based domain?
Thanks,
Colin
Colin