David Klee loops us in on a tricky-to-catch problem:
We’ve been tracking a weird state with SQL Server virtual machines on VMware and possible warnings on database corruption while VM backups are running, largely centered around (but not isolated to) the tempdb database.
TLDR: We’ve now got a VMware KB article on this situation that you and your VM admins should read if you hit the condition and fall into the specifics listed below. Reference VMware KB 88201 for more details.
Read on for David’s thoughts and what to do if you hit this problem.
This seems to be an issue when the OS is being backed up and the snapshot is created at the volume level. If you are taking separate database backups, change the data and log volumes from Persistent to IndependentPersistent. When an OS backup is started, it will not create a snapshot of those volumes–it will just keep a stub for the volume configuration. You will lose bare metal recovery capability, but you’d have to restore the latest backup of the databases anyway.
Thanks for the additional info, Ron.