zfs discuss - Feb 2007 - Zpool complain about missing devices

Hello,
   We had  a situation at customer site where one of the zpool complains about
missing devices. We do not know which devices are missing. Here are the details:

Customer had a zpool created on a hardware raid(SAN). There is no redundancy in
the pool. Pool had 13 LUN''s, customer wanted to increase the size of
and added 5 more Luns. During zpool add process system paniced with zfs checksum
error. After system rebooted, it did not import pool and force import is
complaining of missing devices. It is a potential data loss situation. I know
that there are some bugs which are related to checkup and other related errors
without replication.

I would like to know from community leaders our position on using non-replicated
zpools on hardware raid. I would really apreciate recommended approach when H/W
raid is involved.

Ajat
 
 
This message posted from opensolaris.org

Ajat wrote:
> Hello, We had  a situation at customer site where one of the zpool
> complains about missing devices. We do not know which devices are
> missing. Here are the details:
> 
> Customer had a zpool created on a hardware raid(SAN). There is no
> redundancy in the pool. Pool had 13 LUN''s, customer wanted to
> increase the size of and added 5 more Luns. During zpool add process
> system paniced with zfs checksum error. After system rebooted, it did
> not import pool and force import is complaining of missing devices.
Is it possible that you mistyped the ''zpool add'' command and
listed the
same device twice?  If so, you are probably hitting 6414648 "zfs allows 
overlapping devices to be added", which was fixed in snv_56.
> I would like to know from community leaders our position on using
> non-replicated zpools on hardware raid. I would really apreciate
> recommended approach when H/W raid is involved.
ZFS will work well -- better than any other filesystem -- on hardware 
RAID even without using replication at the zpool level.  There are 
additional benefits (beyond those offered by other filesystem / volume 
manager / hardware RAID) available to you by doing some replication at 
the zpool level (namely, more types of hardware errors can be corrected).

--matt