Re: zfs corruption at zroot/usr/home:<0x0>

On Thu, Nov 13, 2025 at 8:54 AM Sad Clouds <cryintothebluesky@gmail.com> wrote:
> On Wed, 12 Nov 2025 20:18:39 +0100
> Tomek CEDRO <tomek@cedro.info> wrote:
>
> > Hmm this is brand new NVME drive not really likely to fail. I have the
> > same problem on zraid0 (stripe) array while initially I saw the bad
> > file name with 3 problems (vm image) it now turned into
> > ztuff/vm:<0x482>. Charlie Foxtrot :-(
>
> Personally I still prefer the hardware RAID. For years I've used second
> hand LSI 9260-8i cards I bought on ebay and not noticed any corruption
> issues.
>
> ZFS has nice features like checksumming and snapshots, however if I
> need ZFS then I deploy it on top of a hardware RAID virtual drive. I
> know people will frown up this configuration, but if I notice any issues
> with ZFS, I can easily switch to UFS and keep the benefits of the
> hardware RAID.
>
> I could be wrong, but I sort of feel the ancient firmware on a hardware
> RAID card is more stable than the large and complex codebase of ZFS
> that is constantly refactored and improved by many people.

Yes hardware RAID will be faster than ZFS I experimented with that
some years back when I played with ZFS for the first time. RAID-0
(stripe) have no backup data built-in so its only faster than ZFS
stripe but prone to data loss too. Higher RAID levels will protect
data at the cost of available space and some performance. But ZFS
gives you a lot more (see below).

Imagine you are making a hardware upgrade with 16TB array. There is
only a slight change it will work on a new hardware unless you also
move the RAID controller. What if the array is even bigger? Where will
you store backups? It will probably cost you double of the disks price
and the time to transfer data. Also every controller has limited
amounts of ports for disks.

With ZFS RAID you can move the array to any hardware, and then
add/replace disks from other controllers, so you are not bound by the
hardware limitations. Recently I moved to a new machine and things
just worked out-of-the-box with zero additional work I was surprised
to be honest!!

ZFS gives you then far more features than any other controller /
filesystem, including creating dedicated datasets with specific
attributes like compression or encryption, deduplication, snapshots,
quotas, backup export/import to any stream over any medium, etc etc.

I just realized my RAIDZ2 has double parity scheme that equals RAID-6
not RAID-5 as I said before sorry for that (RAID-5 ~ RAIDZ). Now I can
add some additional disks with no problem. And it looks self-healing
true because two other simple stripes detected data corruption while
raidz2 did not. And the corruption affects only one dataset not the
whole pool so I can either delete questionable files and restore them
or roll back snapshot / export for that specific single dataset not
the whole pool. If you think about lost efficiency then you can enable
compression then for HDD things work faster not slower (kind of
surprise too because amount of data written is smaller and the
compression is ultra fast).

I found this article on comparing RAID-Z with RAID helpful:

https://www.diskinternals.com/raid-recovery/what-is-raidz/

ZFS is really amazing and makes FreeBSD unique, even OpenBSD does not
have that, so for people avoiding Linux this helps making the BSD
choice easier :-P

--
CeDeROM, SQ7MHZ, http://www.tomek.cedro.info