mjhjmtu

I know that ZFS properties like copies and compression only affect newly written data.

However I wonder if a scrub would update that?

Let's say that I have have created a pool and set compression=lz4 and copies=2 before writing 1TB of files.

Then I decided I don't need to keep the ditto blocks, but I also would like to use another type of compression.

If I now set copies=1 and compression=gzip-9, is there a way to apply this to data that is already written to the pool?

Would a scrub do that for me?

No.

Changing dataset properties like compression and copies only affects files written after the change. To apply changes like this to existing files, you would need to copy them and mv them over the original. This will, of course, break any connection to any prior snapshots of the same filename (and also to any hard links to the file as the inode will be different).

Alternatively, to apply such changes to an entire pool or dataset, you could zfs send a snapshot to a different pool (e.g. a backup pool), destroy the dataset from the original pool (or destroy the pool and re-create it), and then zfs send it back. Note: you can not do this with zfs send's -R (--replicate) option because that also turns on send's -p (--props) option. See man zfs and search for zfs send for more details.

zfs scrub checks the existing data on a pool and rewrites any corrupted copies if there is sufficient redundancy to have a good copy that matches the checksum...if not, it just warns of the un-correctable error.