[zfs-discuss] Big mistake adding disks to zpool :-(

Gregor Kopka zfs-discuss at kopka.net
Mon Dec 15 03:20:00 EST 2014


On your mappings you had write errors, so it well could be that the
initial writes (which should have created the labels) didn't make it to
disk in the first place.

If you're sure that all devices are reachable through the mappings you
gave to ZFS you could try to issue a /zpool clear/ to see if you can get
the pool unstuck.

Gregor

Am 14.12.2014 um 22:05 schrieb Smithies, Russell:
>
>  
>
> Yep, the “good” disks display the labels correctly eg.
>
>  
>
> gibraltar# zdb -l /dev/sdb1 | less
>
> --------------------------------------------
>
> LABEL 0
>
> --------------------------------------------
>
>     version: 5000
>
>     name: 'z301'
>
>     state: 0
>
>     txg: 2132334
>
>     pool_guid: 1218843039939203747
>
>     errata: 0
>
>     hostname: 'granaspp01.agresearch.co.nz'
>
>     top_guid: 8194404849776417989
>
>     guid: 7294925818686282342
>
>     vdev_children: 23
>
>     vdev_tree:
>
>         type: 'raidz'
>
>         id: 0
>
>         guid: 8194404849776417989
>
>         nparity: 2
>
>         metaslab_array: 46
>
>         metaslab_shift: 36
>
>         ashift: 12
>
>         asize: 12002308128768
>
>         is_log: 0
>
>         create_txg: 4
>
>         children[0]:
>
>             type: 'disk'
>
>             id: 0
>
>             guid: 7294925818686282342
>
>             path: '/dev/disk/by-vdev/G1-part1'
>
>             whole_disk: 1
>
>             DTL: 3147
>
>             create_txg: 4
>
>         children[1]:
>
>             type: 'disk'
>
>             id: 1
>
>             guid: 8924724070060668096
>
>             path: '/dev/disk/by-vdev/G2-part1'
>
>             whole_disk: 1
>
>             DTL: 3146
>
>             create_txg: 4
>
>         children[2]:
>
>             type: 'disk'
>
>             id: 2
>
>             guid: 6473665244239287366
>
>             path: '/dev/disk/by-vdev/G3-part1'
>
>             whole_disk: 1
>
>             DTL: 3145
>
>             create_txg: 4
>
>         children[3]:
>
>             type: 'disk'
>
>             id: 3
>
>             guid: 14675902168276287335
>
>             path: '/dev/disk/by-vdev/G4-part1'
>
>             whole_disk: 1
>
>             DTL: 3144
>
>             create_txg: 4
>
>         children[4]:
>
>             type: 'disk'
>
> :
>
>  
>
> But the dud disks don’t:
>
>  
>
> gibraltar# zdb -l /dev/sdct1
>
> cannot open '/dev/sdct1': No such file or directory
>
> gibraltar# zdb -l /dev/sdct
>
> --------------------------------------------
>
> LABEL 0
>
> --------------------------------------------
>
> failed to unpack label 0
>
> --------------------------------------------
>
> LABEL 1
>
> --------------------------------------------
>
> failed to unpack label 1
>
> --------------------------------------------
>
> LABEL 2
>
> --------------------------------------------
>
> failed to unpack label 2
>
> --------------------------------------------
>
> LABEL 3
>
> --------------------------------------------
>
> failed to unpack label 3
>
>  
>
> I think the dud disks have old labels as they were (probably) part of
> an old test pool and I didn’t clear them before adding to the new pool.
>
>  
>
> Starting on “Plan B”, copying off the data (82TB) to an adjacent
> server then I’ll rebuilt the pool, copy the data back then re-sync the
> replication.
>
> Should all be completed by the time I return from Christmas break!
>
>  
>
> --Russell
>
>  
>
>  
>
> *From:*Bernd P. Ziller [mailto:bernd.ziller at jaxu.de]
> *Sent:* Monday, 15 December 2014 9:14 a.m.
> *To:* Smithies, Russell
> *Subject:* Re: [zfs-discuss] Big mistake adding disks to zpool :-(
>
>  
>
>  
>
> Am 14.12.2014 um 20:56 schrieb "Smithies, Russell"
> <Russell.Smithies at agresearch.co.nz
> <mailto:Russell.Smithies at agresearch.co.nz>>:
>
>     The mappings are correct, I checked with fdisk and matched up the
>     WWNs first.
>
>     The drives are physically there, this is drive I33 which maps to
>     /dev/sdnn
>
>  
>
> Have you already checked the mappings with with 'zdb -l'?
>
> That many unavailable disks means either massive hardware problems or
> a reordering of your drives.
>
>  
>
>  
>
>  
>
> To unsubscribe from this group and stop receiving emails from it, send
> an email to zfs-discuss+unsubscribe at zfsonlinux.org
> <mailto:zfs-discuss+unsubscribe at zfsonlinux.org>.

To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe at zfsonlinux.org.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.zfsonlinux.org/pipermail/zfs-discuss/attachments/20141215/9808ff60/attachment.html>


More information about the zfs-discuss mailing list