[OpenIndiana-discuss] zpool in sorry state

[Eric Pierce]

Indeed, right now zpool status -v is reporting only 1 unrecoverable error.
However, other LUNs aren't recognized by VMWare as VMFS volumes anymore.

The server does have ECC memory, and an LSI SAS controller (no RAID, ZFS
handles everything).  We've had this in production for about 4 months
without issue until yesterday.  I agree about mirror-8; I'm concerned it may
also have problems.  I did online one drive in mirror-3.  The other drive,
wouldn't online.

At this point I want to get the two replacement drives I have installed,
which goes back to one of my questions:  what's the best method for
replacement with hot spare already in place?  I've seen other articles/posts
about simply using zpool detach pool_name failed_device, but I want to make
sure I get that part right and don't cause further problems.  Once I know
those are in place, I'm going to run a full scrub in the evening.

Thanks,
Eric Pierce

On Fri, Jul 8, 2011 at 9:27 AM, Lucas Van Tol <catseyev9 at hotmail.com> wrote:

>
> I think the re silver should have looked at all the data and given you the
> entire list of bad data, but I'm not entirely sure if re silvers look
> outside of the vdev they are fixing.
> A scrub would look at all the data and verify it.
>
> I note that your drives are out due to too many errors.
> Normally, I would say just replace them, but since you lost so many at
> once; it may be worth trying to force the drives online.
> I don't know how that will interact with the spares, but if the original
> drives can be brought online and a scrub run; one of the two drives may
> still have the missing data; after which a scrub may recover it.
> It would be a problem if the spare detached itself though, and I'm not sure
> how spares behave in such situations.
>
>
> Also; mirror 8 has more 'read' errors than I would trust.
> I find it suspicious that there are an equal #'s of failures on disks in
> that vdev.
> Have you had any memory or disk controller issues on the system, and are
> you using ECC memory?
>
>
>
> -Lucas Van Tol
>
>
>