[TriLUG] Disk drive recovery services

[Joseph Mack NA3T via TriLUG]

On Tue, 22 Sep 2020, Wes Garrison via TriLUG wrote:

Hi Wes,

> It wasn't the fault of ZFS.

ah.

> I was using the notorious Seagate ST3000DM001, which is a consumer drive not 
> meant for NAS (not meant for much of anything really).

OK

> Out of the original 8, only 2 remain.  Worst failure rate I've ever seen.

:-(

> I was using a ZFS Z2 array (which can support 2 simultaneous failures), and 
> had 3 simultaneous failures; the latter 2 occurring during re-silvering. This 
> is pretty common, apparently.  When you have a failure and start hammering the 
> drives to do a rebuild, you often have more failures.

Jim Salter (gave us a talk about 2 years ago), says to only use mirrors to avoid 
rebuild problems. Here's why you shouldn't RAID anymore (anymore started in 
2015). Instead just mirror.

https://jrs-s.net/2015/02/06/zfs-you-should-use-mirror-vdevs-not-raidz/

Also it seems crazy, now that he points it out, to have to access 5 drives (or 
whatever) to get your read or your write.

> Lessons learned:
> 1) Use NAS drives or Seagate Exos for RAID/ZFS arrays, not crappy consumer
> drives.  If you buy WD, make sure you get CMR and don't get fleeced with
> SMR drives.

yup.

> 2) use ZFS Z3 for better fault tolerance

Jim says to mirror. After reading his article, I'm convinced.

> 3) have an actual backup if practicable.  This is hard when your array is 
> huge.  How do you back up an 18TB array when the largest available consumer 
> drive is 4TB?  With another 18TB array, of course!

That's how we did it when we had 10MB drives too.

Thanks Joe

-- 
Joseph Mack NA3T EME(B,D), FM05lw North Carolina
jmack (at) wm7d (dot) net - azimuthal equidistant
map generator at http://www.wm7d.net/azproj.shtml
Homepage http://www.austintek.com/ It's GNU/Linux!