Migration from symlink/hardlink to zfs/btrfs datasets


I am experiencing a lot of slowness since the last version (seperate post), and in general the hardlink/slowlink method seems to need to lot of overhead/maintenance.

So i wonder if it would be better to use the new storage mode that can take advantage of btrfs/zfs.
The storage is already zfs based, just it doesn’t seems to use the new mode, i didn’t created the dataset file.

Do i need to initiate a storage migration ? by using /etc/urbackup/dataset , by using migrate_storage_to ?

Would the existing backup actually be migrated or only the new ones will be stored on the new backend?

I think there were some issue with a very large inode file, is it linked to this kind of migration?
As i remember the inode file was roughly doubling the required storage, is it still the case, or can it migrate one client, delete the inode file, then another client , etc… ?

bump @uroni

Got some issues that i highly suspect to be linked to zfs.
So, can i migrate all the data to btrfs, and migrate existing data to the new storage mode, if so, how, whats the requirement, best practice?