I have a parent dataset under which are stored a number of other datasets that are still in use. In the past I also used the parent itself to store data, as it was part of a directory hierarchy.
However, I no longer need the parent dataset's contents, I only need it for organisation (for the children to be categorised under, inherit settings from etc.).
What is the best option for fully cleaning the parent dataset?
In the past when I've done something like this I simply mounted the parent on its own, and deleted everything from it, but I found that ZFS was still showing 200mb as referenced and logicalreferenced despite there being no trace of anything in the dataset (including hidden files). Another dataset I did this to only had about 17mb, but both had a similar number of child datasets, which struck me as odd – what might ZFS still be referencing in the larger of the two?
Neither of these have snapshots (except temporarily for sending, I used bookmarks to retain a point for sending from). Is usedbydataset affected by the size of children? I wouldn't expect it to be, but the "cleared" parent with the largest referenced value does have by far the largest children of the two.
Is there some other way to "clean" a dataset so that it contains nothing except for its properties, and the bare minimum required for the children?