mdadm RAID problem -- won't boot

[Tom H]

On Mon, Aug 20, 2012 at 3:39 PM, Rick Bragg <rbragg at gmnet.net> wrote:
>> On Mon, Aug 20, 2012 at 1:36 PM, Rick Bragg <rbragg at gmnet.net> wrote:
>>>
>>> I am having a problem booting my system.  My boot disk is not a raid array,
>>> however, I do have 4 other disks making a raid 10 array that I mount at /mnt/md0.
>>> My problem is that when I boot my system, I get to a point where it says it
>>> it can't start the degraded array, and asks me if I want to start the degraded
>>> array.  If I say yes or no, it always drops me to a shell.  At the shell, I do a
>>> "cat /proc/mdadm" and I can see 2 arrays!  One is /dev/md0 started, degraded with
>>> only 3 of my disks (sda1, sdc1, sdd1.)  The other array is /dev/md127 with the
>>> other disk all by itself (sdb1) and not started.  Again, I am booting from a
>>> different disk entirely (sde1.)  I tried to remove the md127 array altogether,
>>> and
>>> re-add sdb1 into the md0 array, and it syncs up fine.  After syncing and seeing
>>> that the md0 array is fine, I reboot.  After rebooting, I get the same problem
>>> over
>>> and over again.
>>> My question is:
>>> How can I fix this so that I only have one array at /dev/md0 with all 4 disks
>>> synced?  Also, how can I bypass this and boot my system without any raid at all
>>> so
>>> I can fix that later?  I am using ubuntu server 10.04 LTS.
>>
>> The use of md127 usually means that the array's recognized as a
>> foreign array. Does "mdadm --examine" on sda1 and sdb1 return the same
>> "local to host" value on the "Name" line?
>>
>> Did you zero the superblock before re-adding sdb1?
>>
>
> I didn't zero the superblock on anything.  Should I zero it on sdb1?
>
> for --examine, I'm seeing:
> RaidDevice       State
> 0                active sync      /dev/sda1
> 1                faulty removed
> 2                active sync      /dev/sdc1
> 3                active sync      /dev/sdd1
> 4                spare            /dev/sdb1
>
> Not sure why sdb is a spare, and dev 1 is faulty removed...
>
> Also, for mdadm --detail /dev/md0 I see:
>
> RaidDevice       State
> 0                active sync        /dev/sda1
> 1                spare rebuilding   /dev/sdb1
> 2                active sync        /dev/sdc1
> 3                active sync        /dev/sdd1
>
> After the spare rebuilds and is active sync like the others, I usually reboot and I
> get the same problem all over from the start.  Do I need to somehow re-create the
> entire array? or zero a superblock somewhere after this syncs?

I've never seen output from "--examine" like that. It looks like the
first's from when sdb1 is alone in md127 and the second's from when
sdb1 is added to md0 and is resyncing.

Once sdb1 is neither in md0 nor in md127, run "mdadm --force
--zero-superblock /dev/sdb1" and add sdb1 to md0.