Related issue
https://github.com/longhorn/longhorn/issues/2897
Test Step
Given with Longhorn v1.2.0-rc2 or above.
And deploy compatible engine image oldEI older than test version (for example: longhornio/longhorn-engine:<previous feature/patch release version>).
And create volume vol-old-engine.
And attach volume vol-old-engine to one of a node.
And upgrade volume vol-old-engine to engine image oldEI.
When create backup of volume vol-old-engine.
Then watch kubectl kubectl -n longhorn-system get backups.longhorn.io -l backup-volume=vol-old-engine -w.
And should see only one backup be left after a while.