zpool replace 'stuck'

[Rich]

Oh, what a horrid fix!

We modified src/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c:~3577 from:
                VERIFY(spa_scrub(spa, POOL_SCRUB_RESILVER) == 0);
to:
                spa_scrub(spa, POOL_SCRUB_RESILVER);

and ~3469:
        if (type == POOL_SCRUB_EVERYTHING &&
to:
        if ((type == POOL_SCRUB_EVERYTHING || type == POOL_SCRUB_RESILVER) &&

This allowed the scrub to finish sanely.

Why is it sane to allow it to abort and start a new resilver when in
mid-resilver?

- Rich

On Mon, Sep 28, 2009 at 2:20 PM, Rich <rincebrain at gmail.com> wrote:
> Hello, world.
>
> On 8.0-RC1 amd64, I seem to have run into the same wedge condition
> noted in http://markmail.org/message/khini6ifoty2ecfd.
>
> Unfortunately, however, I can't use the workaround of removing the
> drive, aborting the scrub, and detaching one of the replacements, as
> it reports thus:
> [root at manticore ~]# zpool detach bukkit 7303939385138290847
> cannot detach 7303939385138290847: no valid replicas
> [root at manticore ~]# zpool detach bukkit da14
> cannot detach da14: no valid replicas
>
> da14 is offline (physically not attached), and has been for several reboots.
>
> Thoughts?
>
> - Rich
>
> --
>
> Os amigos são a forma de Deus cuidar de nós.
>



-- 

As senadoras? Sim! Missa, roda nessa! -- palíndromo