This was from my old blog entry from way back November 2006 and remembered how easy it was to fool SQL Server when trying to replace disks. I was configuring a clustered SQL Server 2008 on a Windows Server 2008 and testing the disk resources for failover. I tried failing over one of the disks and it went out pretty well. The next disk resource that I failed over happened to be the one containing the system databases. To my surprise, it did not failover to the other node causing the SQL Server resource to fail as well. It took me quite a while figuring out that the disk subsystem is somehow corrupted (good thing it was just a test environment) so what I ended up doing was recreating a new disk subsystem and redefining it on my cluster using Failover Cluster Management. I made it available on my SQL Server 2008 resource. I knew I would definitely have downtime as I need to bring down the SQL Server cluster resource. Here's what I did - I brought the SQL Server resource offline, copied the contents of the original disks that contain my system databases to the new disk resource, changed the drive letter of the original disk to something else and rebooted the node that originally hosted the disk resource that is failing. Once it is up, I can reuse the same drive letter to the new disk resource which now has the original system databases. When I brought te SQL Server resource back online, it's as if nothing had happened (except, of course, for some errors in the Clustering error log).
Note that this is definitely not a substitute to a valid backup and restore process for your SQL Server databases. All I'm saying is it just works