Re: [Fwd: Re: Missing superblock on one of the raid devices on raid 0 with 1.2 metadata]

[Benjamin ESTRABAUD <be@mpstor.com>]

On 15/03/13 08:24, Ivan Yordanov wrote:
> Hi Phil,
Hi Ivan,
> The detailed info for this broken raid 0 is here:
>
> mdadm --version
> mdadm - v3.1.4 - 31st August 2010
>
> mdadm -E /dev/sda1
> mdadm: No md superblock detected on /dev/sda1.
Provided that *only* the metadata was zeroed, you should be able to get 
a full recovery of your array.
> mdadm -E /dev/sdb1
> /dev/sdb1:
>            Magic : a92b4efc
>          Version : 1.2
>      Feature Map : 0x0
>       Array UUID : 314351c1:8fd287cb:21bd5e93:56aa71c7
>             Name : gandalf:0  (local to host gandalf)
>    Creation Time : Mon Aug 23 13:47:57 2010
>       Raid Level : raid0
>     Raid Devices : 4
>
>   Avail Dev Size : 312574594 (149.05 GiB 160.04 GB)
>      Data Offset : 2048 sectors
>     Super Offset : 8 sectors
>            State : clean
>      Device UUID : 4a303267:fd84a859:ecb68475:a797a615
>
>      Update Time : Mon Aug 23 13:47:57 2010
>         Checksum : 2739d1d7 - correct
>           Events : 0
>
>       Chunk Size : 512K
>
>     Device Role : Active device 1
>     Array State : AAAA ('A' == active, '.' == missing)
>
> mdadm -E /dev/sdc1
> /dev/sdc1:
>            Magic : a92b4efc
>          Version : 1.2
>      Feature Map : 0x0
>       Array UUID : 314351c1:8fd287cb:21bd5e93:56aa71c7
>             Name : gandalf:0  (local to host gandalf)
>    Creation Time : Mon Aug 23 13:47:57 2010
>       Raid Level : raid0
>     Raid Devices : 4
>
>   Avail Dev Size : 312574594 (149.05 GiB 160.04 GB)
>      Data Offset : 2048 sectors
>     Super Offset : 8 sectors
>            State : clean
>      Device UUID : 72c265f4:76e8edcc:7154fd89:77478688
>
>      Update Time : Mon Aug 23 13:47:57 2010
>         Checksum : af0b14cf - correct
>           Events : 0
>
>       Chunk Size : 512K
>
>     Device Role : Active device 2
>     Array State : AAAA ('A' == active, '.' == missing)
>
> mdadm -E /dev/sdd1
> /dev/sdd1:
>            Magic : a92b4efc
>          Version : 1.2
>      Feature Map : 0x0
>       Array UUID : 314351c1:8fd287cb:21bd5e93:56aa71c7
>             Name : gandalf:0  (local to host gandalf)
>    Creation Time : Mon Aug 23 13:47:57 2010
>       Raid Level : raid0
>     Raid Devices : 4
>
>   Avail Dev Size : 312574594 (149.05 GiB 160.04 GB)
>      Data Offset : 2048 sectors
>     Super Offset : 8 sectors
>            State : clean
>      Device UUID : e5f46d32:b2e0f6b4:0b361c98:9689dcb2
>
>      Update Time : Mon Aug 23 13:47:57 2010
>         Checksum : d916338 - correct
>           Events : 0
>
>       Chunk Size : 512K
>
>     Device Role : Active device 3
>     Array State : AAAA ('A' == active, '.' == missing)
The fact that you have the position of all the other drives from the 
array is good. Now we want the last drive's superblock to be written. 
Since we know the position of all the drives, and assuming you know the 
*exact* arguments passed to mdadm when you first created your raid0 
(correct metadata version, chunk size, etc. (most can be found in the 
existing superblocks), you could call "mdadm --create " with the same 
version of mdadm and MD used when creating the array initially, the same 
options and arguments, and *very important* the drives in the same 
order, which I believe to be: /dev/sda1 /dev/sdb1 /dev/sdc1 /dev/sdd1 
(according to the info above).

This will create a new array, but since you are recreating the same 
*exact* array, the existing data should be there and available untouched.

However, as a word of warning, many things can go wrong this this 
command: If you were to recreate the array slightly differently and 
start overwriting your array you would destroy the data on it. The fact 
that it is a RAID0 is good since creating a new array won't start a 
resync that could be fatal should you have made a mistake providing the 
arguments for the recreation. So the above should be generally safe, 
provided you keep a copy of the information you gave us above and match 
the "create" arguments perfectly.
>
> uname -a
> Linux gandalf 3.5.0vs2.3.4-vs2.3.4 #1 SMP Tue Jan 8 08:31:15 EET 2013
> x86_64 Intel(R) Core(TM) i3 CPU 530 @ 2.93GHz GenuineIntel GNU/Linux
>
> Thanks for your help
> Ivan Yordanov

Regards,
Ben.
>
> On Fri, 2013-03-15 at 10:19 +0200, Nikolay Kichukov wrote:
>> email message attachment, "Forwarded message - Re: Missing superblock
>> on one of the raid devices on raid 0 with 1.2 metadata"
>>> -------- Forwarded Message --------
>>> From: Phil Turmel <philip@turmel.org>
>>> To: Nikolay Kichukov <nkichukov@taxback.com>
>>> Cc: linux-raid@vger.kernel.org, Ivan Yordanov
>>> <iyordanov@taxback.com>
>>> Subject: Re: Missing superblock on one of the raid devices on raid 0
>>> with 1.2 metadata
>>> Date: Thu, 14 Mar 2013 11:32:47 -0400
>>>
>>> On 03/14/2013 10:27 AM, Nikolay Kichukov wrote:
>>>> Hi all,
>>>>
>>>> We are trying to recover a broken raid 0. It consisted of 4 raid
>>>> devices. One of them got the metadata/superblock zeroed and now the raid
>>>> cannot assemble.
>>>>
>>>> OS: Gentoo Linux
>>>>
>>>> My colleague will be able to provide more information regarding kernel
>>>> version and mdadm version.
>>>>
>>>> Is there a way to copy the superblock/metadata from one of the remaining
>>>> drives and edit it prior to placing it on the zeroed drive so that the
>>>> raid can be assembled?
>>>>
>>>> Any hints and pointers are welcomed. Is it possible to fix the raid in
>>>> the first place? If yes, then how do we locate the superblock/metadata
>>>> that needs to be copied and edited from one of the raid member devices?
>>> Start by showing the output of "mdadm -E /dev/sdX" for all of the other
>>> member devices or partitions.  It's likely to be possible to fix your
>>> problem.
>>>
>>> Phil
> --
> 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
>