Hacker News Reader

Other advantages:

-Unlike btrfs subvolumes, zfs datasets can be mounted with different zfs-specific mount options, such as compression algorithms or recordsizes.

-zfs can take atomic recursive snapshots of nested datasets, whereas btrfs snapshots of a subvolume do not include nested subvolumes.

Overall, zfs treats datasets as first-class citizens, whereas the only purpose of btrfs subvolumes seems to be to exclude folders from snapshots.

> With that said, ZFS on linux is slightly awkward as it's out of tree, and most distros build the module with DKMS. I don't entirely trust this for using for / (so my / is just a md raid1) - I use zfs for bulk data instead. btrfs is in-kernel, so there's no real disadvantage with using it for /.

Is that true? RHEL/CentOS have both DMKS and kmod versions; in Ubuntu ZFS is a supported package. That's not most distros, certainly, but it is the ones most people use.

Based on using CentOS + DKMS, Debian + DKMS, and Arch + DKMS :) Some of these have prebuilt modules too, but as community projects they usually lag behind kernel releases to the main distro, and that can be a problem. So I always end up using DKMS.

Not used Ubuntu ZFS, admittedly. Haven't used Ubuntu since 2010 or earlier.

> zfs's logical volume layer is rather more flexible than btrfs's - you can make a multi-device set of mixed disks. For example, my backup pool is 9x4TB and 7x3TB.

You may wish to checkout ZFS dRAID, which recently got committed:

* https://www.youtube.com/watch?v=jdXOtEF6Fh0

> - zfs's logical volume layer is rather more flexible than btrfs's - you can make a multi-device set of mixed disks

Can you please elaborate on this? Working on mixed disks was always a killer feature of btrfs for me!

I do a pool called backup with two raidz vdevs - 7x3T in one raidz vdev, and 9x4T in one raidz vdev. Data is striped on each of the two raidz and I lose one 3T and one 4T's worth of capacity. This allows me to lose any one drive at a given time without data loss, and up to two drives (as long as they come out of separate vdevs).

As far as I could work out from the btrfs documentation, this isn't currently possible? Plus, RAID5/6 in btrfs is still of questionable stability?

Thanks for clarification!

Indeed, btrfs doesn't allow such configuration. However, it allows using disks of unequal sizes within single filesystem - like, 3+3+4T in RAID1 mode gives you 5T of usable space (imagine 4T disk being split in half, and each half duplicated to a different 3T disk; remaining space of 3T disks duplicated to each other). But as I understand, it's possible to achieve the same with manual partitioning and vdev allocation on ZFS, too.

RAID5/6 is indeed a danger zone - in another thread I've already mentioned a nice write-up of "guidelines for users running btrfs raid5 arrays to survive single-disk failures without losing all the data" by Zygo Blaxell: https://lore.kernel.org/linux-btrfs/20200627032414.GX10769@h...