[zfs-discuss] Shuffling disks around a 2 x mirror pool

Gordan Bobic gordan.bobic at gmail.com
Mon Sep 2 07:47:08 EDT 2013


When I had the same problem, it was caused by the partition table zfs
created on the new disk it tried to add being different from the one a
previous version created on the existing disks. Specifically, the newer
version creates a smaller partition 1, relative to the size of the disk. So
when you go to replace a disk with a newer version, it creates smaller
part1, then tries to add it, and bails saying it's too small. Create the
partitions manually so they are exactly the same as on the old disk, and
re-add by partition, and everything works fine.

Regardless of whether you are referencing by whole disk or part1, you
always end up using part1. If you create the pool based on pre-created
partitions, it will use the partitions as is (without creating a partition
table on the partition).




On Mon, Sep 2, 2013 at 12:38 PM, Ryan How <ryan at zbit.net.au> wrote:

> Hi,
>
> Thanks. I have noticed the partitions that ZFS makes. I did try wiping out
> the partition table for it to create a new partition table, but it didn't
> seem to work. Perhaps it was another issue though.
>
> I wonder, if it is referencing it by partition instead of the full disk, I
> wonder if it is then controlling the disk correctly, setting the noop
> scheduler and so forth. Also if it sees the existing partition table, why
> not clobber it and recreate or just use the existing one rather than making
> it reference the partition and giving a strange kernel error?
>
> Anyway, it is working now, so I'm happy :). It's just so ugly seeing
> faulted, degraded, etc all over my pool. Hopefully the re silver makes it
> through. I've had 3 failed disks (out of 6 in the batch that was purchased)
> in a week. All 18 months old. Scary...
>
> Thanks, Ryan
>
>
> On 2/09/2013 5:37 PM, Niels de Carpentier wrote:
>
>> I don't get it?. I ALWAYS use full disks, howcome now I have to use a
>>> partition?
>>>
>> If you use a full disk, zfs will still create a partition table, with
>> partitions 1 and 9. partition 9 is a small 8MB partition for padding, to
>> handle slight changes in disk size in case of replacement.
>>
>> If you look at your other disks, you will see have a partition table as
>> well.
>>
>> I guess the replace with /dev/sdb would have worked if you removed the
>> partition table first. (So zfs would create the partition table used for
>> full disks). This might be considered a bug, I guess the best thing would
>> be for zfs to show a clear error message if you try use a full disk with
>> an existing partition table.
>>
>> Niels
>>
>>
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to zfs-discuss+unsubscribe@**zfsonlinux.org<zfs-discuss%2Bunsubscribe at zfsonlinux.org>
>> .
>>
>
>
> To unsubscribe from this group and stop receiving emails from it, send an
> email to zfs-discuss+unsubscribe@**zfsonlinux.org<zfs-discuss%2Bunsubscribe 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/20130902/4974b62e/attachment.html>


More information about the zfs-discuss mailing list