ZFS with errors

[Brandon J. Wandersee]

Luciano Mannucci writes:

> On Wed, 13 Apr 2016 15:56:20 +0200 (CEST)
> Trond Endrestøl <Trond.Endrestol at fagskolen.gjovik.no> wrote:
>
>> There's no redundancy in this pool, making it hard for ZFS to 
>> automatically repair your files.
>> 
>> Maybe you should destroy your pool and recreate it using a mirrored 
>> configuration. Maybe, mirror disks 1 & 2, and disks 3 & 4, e.g.
> They are of different sizes. I don't know if I can add redundancy
> without loosing the bits that exeed the smallest one...
> I'm a zfs newbie, just experimenting by now... :)

Experimenting is good, but you've sort of dived head-first into the bad
practices deep end. A striped ZFS pool consisting of four disks of
different sizes is probably *less* reliable (and possibly less
performant) than a a single disk formatted with a traditional
filesystem. Your data has no redundancy, the pool will only perform as
well as the slowest of the four disks, and a problem with any one disk
will affect all data. Even a one-disk ZFS stripe with the "copies"
property raised is probably better than a large, multi-disk stripe.

You can read the original "ZFS Best Practices" guide for some more
info.[1] Some of it is either slightly out-of-date, or applies only to
Solaris, but most of it is still relevant to FreeBSD today.

If you're willing to spend a little money, "FreeBSD Mastery: ZFS" by
Michael W. Lucas is pretty good.[2]

[1]: http://www.solarisinternals.com/wiki/index.php/ZFS_Best_Practices_Guide
[2]: https://www.tiltedwindmillpress.com/?product=fmzfs

-- 

::  Brandon J. Wandersee
::  brandon.wandersee at gmail.com
::  --------------------------------------------------
::  'The best design is as little design as possible.'
::  --- Dieter Rams ----------------------------------