[zfs-discuss] zpool import -a problem

Darik Horn dajhorn at vanadac.com
Tue Dec 13 19:04:37 EST 2011

On Tue, Dec 13, 2011 at 17:44, Ben Hodgens <aequitas at gmail.com> wrote:
> Hi all,
> I recently changed controllers from horrible Marvell controllers
> (which caused almost immediate lockups, I think) to LSI (IBM branded)
> 1068e cards. This is on Debian 6, with root on a separate md0 array on
> the onboard controllers.

What was the model number of the Marvell part?  -- I'm trying to keep
a list of hardware that is known to be bad for ZoL.

> After changing the controllers out for the new ones, I did an zpool
> import -a. I had initially built the array using /dev/disk/by-id.
> However, upon import, it got imported as 16 disks (the correct number)
> of sd# - eg sdc, sdd, sde, and so on. I can see how this might pose a
> problem in identifying a failed disk, for instance.
> I just thought I'd report this inconsistency in the event that it may
> be able to be remedied. (I'd guess the whole 'replace failed disk'
> thing might be a higher priority?)

The import command takes the -d switch too.  You'll get the desired
result if you export the pool and do this:

  # zpool import -d /dev/disk/by-id -a

Darik Horn <dajhorn at vanadac.com>

More information about the zfs-discuss mailing list