Re: Netbsd 5.0RC2 amd64, system raid fail to reconstruct

[Thomas Galliano <maillistthom%googlemail.com@localhost>]

2009/3/23 Manuel Bouyer <bouyer%antioche.eu.org@localhost>:
> On Sat, Mar 21, 2009 at 10:29:37AM +0100, Thomas Galliano wrote:
>> I used jibbed liveCD that use netbsd 5.0RC1 amd64
>> I did the same thing:
>>    - raidctl -A no raid0
>>    - raidctl -u raid0
>>    - dd if=/dev/zero of=/dev/rwd1d bs=8k count=1
>>    - fdisk -0ua /dev/rw1d
>>    - disklabel -R -r disklabel.wd1 wd1
>>    - redef raid using an absent component instead of the one out of
>> sync (raidctl -C)
>>    - raidctl -I 200903201 raid0
>>    - raidctl -A root raid0
>>    - shutdown -r now
>> reboot on the local system Netbsd 5.0RC2 amd64
>>    - raidctl -a /dev/wd1a raid1
>>    - raidctl -F component1 raid1
>
> If I understand it properly, you rebuild the raid on wd1, leaving wd0
> unconfigured. Then your reboot and add wd1 as a hot spare for raid0,
> so you're destroying your just-rebuild raid to use the disk as hot-spare
> for another raid.
>

Not exactly. I got my root FS on this raid, so I preferred not to
loose these data even if their just for personal use.
So I unconfigure the raid. Kept wd0a in the array and mark the second
component as absent.
I put a new Serial Number on this new raid containing wd0a and an
absent component.
Then I reboot and try to hotspare the second hard drive wd1a which
failed to reconstruct.

Here the raid config:
START array
 # numRow numCol numSpare
 1 2 0

 START disks
 /dev/wd0a
absent

 START layout
 # sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_1
 128 1 1 1

 START queue
 fifo 100


> Did you use the same Serial Number for old and new raid ? This may
> be part of the problem.
>
> --

No I didn't. But I used the raidctl -I xxx when I configured the raid
with the absent component.
I never used raidctl -I xxx when wd1a was part of the raid.
Yet the procedure has worked with the other raid :(