I have been working with Veeam software since version 4 and other VM backup applications for many organizations over the years. In the VM DR arena there is always some risk associated with relying on crash-consistent backups -- particularly with databases. I've seen cases where VM restores resulted in databases that ended up in a corrupted state and could not be fully recovered to a consistent state. The response I've seen to this is usually to use the Application Aware Image Processing (AAIP), which incorporates Microsoft VSS to attempt to gain application consistency at the time of backup. Honestly, I have avoided using VSS because I have seen mixed results and it seemed to add additional complexity to the backup process. Some workloads also of course don't support VSS either. That said, however, I am interested in a more comprehensive strategy. I am curious if others have seriously considered the option of performing periodic fully consistent offline backups. I understand the risks with online backups have been partially addressed through other means like recommending DBMS backups to flat files, which I certainly do embrace where possible -- but it is unfortunately difficult to manage and maintain visibility. I wonder though if the offline backup strategy is worth incorporating right into a comprehensive Veeam backups/replication strategy -- for example, administrators may want it for certain workloads once a week, and could schedule downtime for it every week.
Would this be worth adding as a feature for both Veeam backups and replication? The capabilities are already built into Veeam actually... It is used with Veeam QuickMigrations... In a QuickMigration Veeam takes a snapshot, replicates the VM, shuts down the VM, replicates the changes since the initial snapshot, and then powers on the VM at the new location.
Of course there is risk on the other side associated with shutting down a VM and powering it back on... (and QuickMigration can fail in those cases), but for important databases, the offline fully consistent backup could prove extremely valuable.
Another edge use case for this would be avoiding the snapshot disk space and I/O overhead entirely by allowing the VM to be powered off for the duration of the entire backup. For example, this could be useful for small business customers that may not have sufficient free space on a datastore because VM level backups were not planned when the VM host server was deployed.
Thoughts?
Thanks for reading.
References:
https://www.veeam.com/blog/how-to-creat ... ackup.html
https://www.veeam.com/blog/powered-off- ... t-too.html
http://forums.veeam.com/post176793.html