This message was deleted.

I'll write a small script to find the cfg files older than 15 days and delete them.

Yes—if there’s leftover VMBackup metadata in the artifact, the Harvester controller will still sync the VMBackup manifest into Harvester. However, Harvester has an enhancement that checks whether the VMBackup’s associated Longhorn backup is still valid; if it isn’t, Harvester will mark the corresponding VMBackup as an error state.

Yeps, that's exactly my problem. And I end up with more than a thousand "backups.longhorn.io" not found vm backups in the backups screen (and poor etcd)