Re: mdadm 3.1.1: level change won't start

["Kristleifur Daðason" <kristleifur@gmail.com>]

On Mon, Dec 21, 2009 at 11:18 PM, Kristleifur Daðason
<kristleifur@gmail.com> wrote:
> On Mon, Dec 21, 2009 at 10:57 PM, Neil Brown <neilb@suse.de> wrote:
>> On Mon, 21 Dec 2009 03:41:33 +0000
>> Kristleifur Daðason <kristleifur@gmail.com> wrote:
>>
>>> Hi all,
>>>
>>> I wish to convert my 3-drive RAID-5 array to a 6-drive RAID-6. I'm on
>>> Linux 2.6.32.2 and have mdadm version 3.1.1 with the 32-bit-array-size
>>> patch from here: http://osdir.com/ml/linux-raid/2009-11/msg00534.html
>>>
>>> I have three live drives and three spares added to the array. When I
>>> initialize the command, mdadm does the initial checks and aborts with
>>> a "cannot set device shape" without doing anything to the array.
>>>
>>> Following are some md stats and growth command output:
>>>
>>> ___
>>>
>>> $ cat /proc/mdstat
>>> Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5]
>>> [raid4] [raid10]
>>> md_d1 : active raid5 sdd1[6](S) sdc1[5](S) sdb1[4](S) sdf1[1] sde1[0] sdl1[3]
>>>       2930078720 blocks super 1.1 level 5, 256k chunk, algorithm 2 [3/3] [UUU]
>>>       bitmap: 1/350 pages [4KB], 2048KB chunk
>>>
>>> $ mdadm --detail --scan
>>> ARRAY /dev/md/d1 metadata=1.01 spares=3 name=mamma:d1
>>> UUID=da547022:042a6f68:d5fe251e:5e89f263
>>>
>>> $ mdadm --grow /dev/md_d1 --level=6 --raid-devices=6
>>> --backup-file=/root/backup.md1_to_r6
>>> mdadm: metadata format 1.10 unknown, ignored.
>>> mdadm: metadata format 1.10 unknown, ignored.
>>> mdadm level of /dev/md_d1 changed to raid6
>>> mdadm: Need to backup 1024K of critical section..
>>> mdadm: Cannot set device shape for /dev/md_d1
>>> mdadm: aborting level change
>>> ___
>>>
>>>
>>> Three questions -
>>>
>>> 1. What does the stuff about "metadata format 1.10 unknown" mean?
>>> Notice the "super 1.1" vs. "metadata 1.01" vs. "metadata format 1.10"
>>> disrepancy between mdsat, --detail and --grow output.
>>
>> The metadata format .. unknown means that your /etc/mdadm.conf contains
>> something like
>>       metadata=1.10
>>
>>>
>>> 2. Am I doing something wrong? :)
>>
>> Not obviously.
>>
>>>
>>> 3. How can I get more info about what is causing the failure to
>>> initialize the growth?
>>
>> Look in the kernel logs.  e.g.
>>   dmesg | tail -20
>>
>> immediately after the "mdadm --grow" attempt.
>>
>> I just tried the same thing and it worked for me.
>>
>> NeilBrown
>>
>
> Thank you very much for the reply. You were right, mdadm.conf indeed
> contained metadata=1.10. I fixed it, updated the initramfs and
> rebooted.
>
> ---
>
> mdadm --detail --scan now gives:
>
>  sudo mdadm --detail --scan
> ARRAY /dev/md/d1 metadata=1.01 spares=3 name=mamma:d1
> UUID=da547022:042a6f68:d5fe251e:5e89f263
>
> ---
>  I tried the grow command again, and it aborts again. Could it be that
> the device sizes are wrong? I thought I meticulously created exactly
> identical partitions on each of the drives. The command output is:
>
>  sudo mdadm --grow /dev/md_d1 --level=6 --raid-devices=6
> --backup-file=/root/backup.md1_to_r6
> mdadm level of /dev/md_d1 changed to raid6
> mdadm: Need to backup 1024K of critical section..
> mdadm: Cannot set device shape for /dev/md_d1
> mdadm: aborting level change
>
> ---

I figured it out - I just needed to disable the write-intent bitmap by
doing "mdadm --grow --bitmap=none /dev/md_d1". Then the reshape was
startable and is going quite well.

I'm seeing resync speeds of around 30-40 MB/second during the
reshape/relevel. I'm running on 1.5TB SATA drives. Is this speed OK?

Thanks everybody for the help and suggestions!
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html