Error mounting a reiserfs on renamed raid1

Error mounting a reiserfs on renamed raid1

[Clemens Koller]

Hi there.

I am new to this list, however didn't find this effect nor a
solution to my problem in the archives or with google:

short story:
------------
A single raid1 as /dev/md0 containing a reiserfs (with important data)
assembled during boot works just fine:
$ cat /proc/mdstat
Personalities : [linear] [raid0] [raid1]
md0 : active raid1 hdg1[1] hde1[0]
       293049600 blocks [2/2] [UU]

The same raid1 moved to another machine as a fourth raid can be
assembled manually as /dev/md3 (to work around naming conflicts),
but it cannot be mounted anymore:
$ mdadm --assemble /dev/md3 --update=super-minor -m0 /dev/hde /dev/hdg
does not complain. /dev/md3 is created. But
$ mount /dev/md3 /raidmd3 gives:

Jan 24 20:24:10 rio kernel: md: md3 stopped.
Jan 24 20:24:10 rio kernel: md: bind<hdg>
Jan 24 20:24:10 rio kernel: md: bind<hde>
Jan 24 20:24:10 rio kernel: raid1: raid set md3 active with 2 out of 2 mirrors
Jan 24 20:24:12 rio kernel: ReiserFS: md3: warning: sh-2021: reiserfs_fill_super: can not find reiserfs on md3

Adding  -t reiserfs doesn't work either.
So, the renaming or reassembling doesn't work, even if /dev/md3 is present
and mdadm --detail /dev/md3 tells that everything is fine?!


long story:
-----------
There are two machines: an old server and a new server.
Servers are in production environment, so there are no risky things to do.
It's all kernel 2.6.23.12 (will try 2.6.23.14 tomorrow) and
mdadm - v2.6.4 - 19th October 2007
Distribution is CRUX (everything vanilla, similar to LFS)

Old server was has it's root on /dev/hda and data on /dev/md0 which is
an Promise PDC20269 dual ATA raid1 configuration consisting of /dev/hde and /dev/hdg.
Everything is reiserfs and working fine.

This machine should be migrated over to a new server with:
With VIA dual SATA raid1 configuration, three partitions for the system, swap and data.
/dev/sd[ab][123] became: /dev/md[012]
md0 is /
md1 is swap
md2 is data.
everything is ext3!

So I plugged the old PDC20269 with the harddisks to the new machine. During boot,
md complains about a duplicate md0:

Jan 24 20:21:44 rio kernel: md: Autodetecting RAID arrays.
Jan 24 20:21:44 rio kernel: md: autorun ...
Jan 24 20:21:44 rio kernel: md: considering sdb3 ...
Jan 24 20:21:44 rio kernel: md:  adding sdb3 ...
Jan 24 20:21:44 rio kernel: md: sdb2 has different UUID to sdb3
Jan 24 20:21:44 rio kernel: md: sdb1 has different UUID to sdb3
Jan 24 20:21:45 rio kernel: md:  adding sda3 ...
Jan 24 20:21:45 rio kernel: md: sda2 has different UUID to sdb3
Jan 24 20:21:45 rio kernel: md: sda1 has different UUID to sdb3
Jan 24 20:21:45 rio kernel: md: hdg1 has different UUID to sdb3
Jan 24 20:21:45 rio kernel: md: hde1 has different UUID to sdb3
Jan 24 20:21:45 rio kernel: md: created md2
Jan 24 20:21:45 rio kernel: md: bind<sda3>
Jan 24 20:21:45 rio kernel: md: bind<sdb3>
Jan 24 20:21:45 rio kernel: md: running: <sdb3><sda3>
Jan 24 20:21:46 rio kernel: raid1: raid set md2 active with 2 out of 2 mirrors
Jan 24 20:21:46 rio kernel: md: considering sdb2 ...
Jan 24 20:21:46 rio kernel: md:  adding sdb2 ...
Jan 24 20:21:46 rio kernel: md: sdb1 has different UUID to sdb2
Jan 24 20:21:46 rio kernel: md:  adding sda2 ...
Jan 24 20:21:46 rio kernel: md: sda1 has different UUID to sdb2
Jan 24 20:21:46 rio kernel: md: hdg1 has different UUID to sdb2
Jan 24 20:21:46 rio kernel: md: hde1 has different UUID to sdb2
Jan 24 20:21:46 rio kernel: md: created md1
Jan 24 20:21:46 rio kernel: md: bind<sda2>
Jan 24 20:21:47 rio kernel: md: bind<sdb2>
Jan 24 20:21:47 rio kernel: md: running: <sdb2><sda2>
Jan 24 20:21:47 rio kernel: raid1: raid set md1 active with 2 out of 2 mirrors
Jan 24 20:21:47 rio kernel: md: considering sdb1 ...
Jan 24 20:21:47 rio kernel: md:  adding sdb1 ...
Jan 24 20:21:47 rio kernel: md:  adding sda1 ...
Jan 24 20:21:47 rio kernel: md: hdg1 has different UUID to sdb1
Jan 24 20:21:47 rio kernel: md: hde1 has different UUID to sdb1
Jan 24 20:21:47 rio kernel: md: created md0
Jan 24 20:21:48 rio kernel: md: bind<sda1>
Jan 24 20:21:48 rio kernel: md: bind<sdb1>
Jan 24 20:21:48 rio kernel: md: running: <sdb1><sda1>
Jan 24 20:21:48 rio kernel: raid1: raid set md0 active with 2 out of 2 mirrors
Jan 24 20:21:48 rio kernel: md: considering hdg1 ...
Jan 24 20:21:48 rio kernel: md:  adding hdg1 ...
Jan 24 20:21:48 rio kernel: md:  adding hde1 ...
Jan 24 20:21:48 rio kernel: md: md0 already running, cannot run hdg1
Jan 24 20:21:48 rio kernel: md: export_rdev(hde1)
Jan 24 20:21:49 rio kernel: md: export_rdev(hdg1)
Jan 24 20:21:49 rio kernel: md: ... autorun DONE.

short story continues here...
I use the full hd[eg] disks for the raid1 with only a single
partition. The partitions are
$ fdisk -l /dev/hde

Disk /dev/hde: 300.0 GB, 300090728448 bytes
255 heads, 63 sectors/track, 36483 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Disk identifier: 0x1bd3d309

    Device Boot      Start         End      Blocks   Id  System
/dev/hde1               1       36483   293049666   fd  Linux raid autodetect

When I plug the old hd[eg] raid back to be a single raid1 in the system,
everything works fine. So nothing was actually broken.

What did I do wrong?

How can I just change md0 to md3 keeping everything else as is?

Thank you,

Clemens

Re: Error mounting a reiserfs on renamed raid1

[Robin Hill]

[-- Attachment #1: Type: text/plain, Size: 1819 bytes --]

On Fri Jan 25, 2008 at 01:48:32AM +0100, Clemens Koller wrote:

> Hi there.
>
> I am new to this list, however didn't find this effect nor a
> solution to my problem in the archives or with google:
>
> short story:
> ------------
> A single raid1 as /dev/md0 containing a reiserfs (with important data)
> assembled during boot works just fine:
> $ cat /proc/mdstat
> Personalities : [linear] [raid0] [raid1]
> md0 : active raid1 hdg1[1] hde1[0]
>       293049600 blocks [2/2] [UU]
>
> The same raid1 moved to another machine as a fourth raid can be
> assembled manually as /dev/md3 (to work around naming conflicts),
> but it cannot be mounted anymore:
> $ mdadm --assemble /dev/md3 --update=super-minor -m0 /dev/hde /dev/hdg
> does not complain. /dev/md3 is created. But
>
It looks like you should be assembling the partitions, not the disks.
Certainly the mdstat entry above shows the array being formed from the
disks.  Try:
  mdadm --assemble /dev/md3 --update=super-minor -m0 /dev/hde1 /dev/hdg1

> $ mount /dev/md3 /raidmd3 gives:
>
> Jan 24 20:24:10 rio kernel: md: md3 stopped.
> Jan 24 20:24:10 rio kernel: md: bind<hdg>
> Jan 24 20:24:10 rio kernel: md: bind<hde>
> Jan 24 20:24:10 rio kernel: raid1: raid set md3 active with 2 out of 2 
> mirrors
> Jan 24 20:24:12 rio kernel: ReiserFS: md3: warning: sh-2021: 
> reiserfs_fill_super: can not find reiserfs on md3
>
> Adding  -t reiserfs doesn't work either.
>
Presumably the superblock for the file system cannot be found because
it's now offset due to the above issue.

HTH,
        Robin
-- 
     ___        
    ( ' }     |       Robin Hill        <robin@robinhill.me.uk> |
   / / )      | Little Jim says ....                            |
  // !!       |      "He fallen in de water !!"                 |

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: Error mounting a reiserfs on renamed raid1

[Clemens Koller]

Robin Hill schrieb:
> On Fri Jan 25, 2008 at 01:48:32AM +0100, Clemens Koller wrote:
> 
>> Hi there.
>>
>> I am new to this list, however didn't find this effect nor a
>> solution to my problem in the archives or with google:
>>
>> short story:
>> ------------
>> A single raid1 as /dev/md0 containing a reiserfs (with important data)
>> assembled during boot works just fine:
>> $ cat /proc/mdstat
>> Personalities : [linear] [raid0] [raid1]
>> md0 : active raid1 hdg1[1] hde1[0]
>>       293049600 blocks [2/2] [UU]
>>
>> The same raid1 moved to another machine as a fourth raid can be
>> assembled manually as /dev/md3 (to work around naming conflicts),
>> but it cannot be mounted anymore:
>> $ mdadm --assemble /dev/md3 --update=super-minor -m0 /dev/hde /dev/hdg
>> does not complain. /dev/md3 is created. But
>>
> It looks like you should be assembling the partitions, not the disks.
> Certainly the mdstat entry above shows the array being formed from the
> disks.  Try:
>   mdadm --assemble /dev/md3 --update=super-minor -m0 /dev/hde1 /dev/hdg1

Argh, that was too simple. I thought I've tried to assemble the
partitions (/dev/hdx1) too instead of the whole disks but I guess
I was wrong.

A simple
mdadm --assemble /dev/md3 /dev/hde1 /dev/hdg1
did the job

Thank you!

Regards,

Clemens