On Thu, Aug 16, 2018 at 07:56:57AM +0200, Anders Andersson wrote: > On Thu, Aug 16, 2018 at 3:47 AM, Zenaan Harkness <zen...@freedbms.net> wrote: > > Another correction below: > > > >> BTRFS also makes it extremely easy to expand an array if you add more > >> disks, WITHOUT mucking around with LVM and md-raid. Just a simple > >> command (something that ZFS cannot easily do at this time) > >> > >> btrfs device add /dev/sdX /path/to/array > >> btrfs filesystem balance /path/to/array > > > > “as simple as zpool add <pool name> <dev name>” > > > > >From here: > > http://list.zfsonlinux.org/pipermail/zfs-discuss/2018-August/031938.html > > > > Note that ZoL does not yet auto-rebalance (although there is at least > > one script around to do it manually - copying files - and rebalance > > can also be done manually with: zfs send ...; zfs receive ... > > I thought the problem was that you can't actually mix-and-match block > devices of different sizes easily in ZFS, you need to plan ahead. I > don't know the details because I've never used ZFS, but when I read > about it I see a lot of warnings about striping and vdevs and whatnot.
If say you have a 3-drive RAID5, you can add extra drives (as above), and also you can add an extra 3 drives as a separate 3-drive RAID "unit" (vdev). Yes if you do the former, and the new drive(s) is/are larger, the excess space will not be used by that ZFS vdev, but you could also partition the spare space and use it as a separate vdev, or for some other fs completely - plenty of options. > With btrfs you can just throw any block devices together and it will > automatically use whatever it can, restricted to the level of > redundancy you requested. Yes, something like that is my understanding too - BTRFS a little "more fancy" in being able to use/stripe across extra space of new drives - there's probably a wiki on it somewhere for the details...
