@mngrif

I used to run a 30tb data store in zfs at the lab. I chose the doubly-redundant setting RAIDZ2, used commodity disks, and scanned for failures a couple times a week. If all the disks were good, I took a snapshot. If one had failed (which did happen as we got bit by the WD flood victim drives) I did the switcheroo & rebuild first.

The system ran debian off a separate SSD and didn't use the zfs system for swap or anything except data storage. Served it on the network as a samba share.