If I have a ZFS dataset, and I add it to incus as a device with source= is there much difference between this and a volume managed by incus?
Seems like at least manageability takes a hit. Like if you’re using incus snapshots instead of an external zfs snapshot manager. ( optional side q: are there gotchas to watch out for it I use external snapshots? )
Any expected performance difference, or buggier code paths?
If there is a difference such that you would recommend volumes, maybe there should be a way for incus to adopt a zfs dataset as a storage volume? Perhaps an extension of incus storage volume import