I'm trying to test setup backups in my Ceph clusters and playing with a couple options. My current proposal would involve creating full image point-in-time snapshots of the Ceph image then exporting them to a cloud storage provider.
The problem with that is when running the rbs export command; the Ceph images' that get exported have their provisioned space instead of the used space; leading to quite large images when I'm just trying to backup the used space
I'm trying to use rbd export-diff instead in order to backup just the changes; but trying to restore that doesn't restore the image to that point in time.
I was expecting using the export-diff command to export the changes between the last snapshot and the current image to a backup file, make some changes to the current image, then run import-diff to apply the backup file and ensure it's essentially "rollbacks" the image to that point in time. Is that thought process correct?