[SOLVED] mount: unknown filesystem type 'linux_raid_member'

didiw · 01-17-2014, 05:46 AM

Hello,

I have the following problem:
I have a 4-scsi-disk software raid5, which had one failed disk. My replacement disk did not work and then, according to Murphy's law, another disk started to fail from time to time under load.
Since I could not run a backup, I mirrored the raid5 to another disk to be able to access it.
In the meantime I had replaced the failing disk and could rebuild the raid5, but I can no longer mount it. IÄm getting the following error:

Quote:

mount: unknown filesystem type 'linux_raid_member'

To get this mounted again, I tried to use losetup with various offsets, but I only got the error message:

Quote:

mount: unknown filesystem type '(null)'

How can I get my files back, at least to make a backup?

Thanks in advance

Regards

didiw

mostlyharmless · 01-17-2014, 06:58 AM

I've only seen that message if I tried to mount a RAID member, as opposed to building the array, then mounting the appropriate array. Could you post the exact commands that you run?

didiw · 01-17-2014, 04:43 PM

Hi,

Quote:

Originally Posted by mostlyharmless

I've only seen that message if I tried to mount a RAID member,

I too get that message,when trying to mount my raid.

Quote:

Originally Posted by mostlyharmless

as opposed to building the array, then mounting the appropriate array. Could you post the exact commands that you run?

I just tried to mount te array as usual:

Code:

mount /dev/md127 /mnt

The result of running this commnd is the error message:

Quote:

mount: unknown filesystem type 'linux_raid_member'

Regards
didiw

mostlyharmless · 01-17-2014, 05:01 PM

Did you run any diagnostics on the array? It rebuilt ok with no problems I take it? Usually my raid arrays look like /dev/md0, btw.

didiw · 01-18-2014, 03:43 AM

hi,

Quote:

Originally Posted by mostlyharmless

Did you run any diagnostics on the array? It rebuilt ok with no problems I take it?

What diagnostics can I run on the array? It rebuilds OK, except for the missing 4th disk.
I don't know, why mdadm chose /dev/md127, but it worked with that name, when the array had no problems and it works now, at least according to

Code:

mdadm --detail /dv/md127

Can I change the name somehow?

Quote:

Originally Posted by mostlyharmless

Usually my raid arrays look like /dev/md0, btw.

I have other raid1 devices, which look like /dev/mdx.

Regards,

didiw

mostlyharmless · 01-19-2014, 08:48 AM

Yeah, I was thinking of mdadm --detail /dev/md127....
OK, I'm stumped. What exactly did you do with

Quote:

Since I could not run a backup, I mirrored the raid5 to another disk to be able to access it.

didiw · 01-19-2014, 05:33 PM

Quote:

Originally Posted by mostlyharmless

Yeah, I was thinking of mdadm --detail /dev/md127....
OK, I'm stumped. What exactly did you do with

Quote:

Since I could not run a backup, I mirrored the raid5 to another disk to be able to access it.

I created a mirror on top of the raid5 with a logical volume as the second disk, using the followng command:

Code:

mdadm --create -l 1 -n 2 --force /dev/md11 /dev/md127 /dev/vgTest/lvTest127

hoping that the disk would survive the mirror action, but it undfortunately did not. And now I am stuck with an unusable raid5 telling me that it cannot be mounted, because it is of type linux_raid_member.
Where is this information stored and what value should be used to set it back to normal?

Regard, didiw

mostlyharmless · 01-20-2014, 08:13 AM

Well I'm guessing that last command creating the mirror damaged the raid meta-data of the original raid5. Perhaps you can try re-assembling the pieces in to a new raid 5 array with a new mdadm --assemble, that might work. In case someone smarter than me is reading this thread (very likely), there might be some value in posting the output of mdadm --detail /dev/md127 and/or cat /proc/mdstat and perhaps the result of the above mirror creation command, if there was one or if you still remember it.

didiw · 01-21-2014, 08:00 AM

Hi,

I finally managed to get my raid5 back.
To get a copy of my data I copied the raw data to a logical volume using dd and worked on that first. I had run fsck on the raid and the logical volume before without and with specifying the standard superblock backup of 8193, but it did not work. Then I read somewhere about dumpe2fs and found the real locattion of the backup superblock. The command

Code:

fsck.ext4 -b <backup_superblock> <logicalValue>

corrected quite a lot of free block and bitmap errors. Afterwards I could mount the logical volume without error message and all my important files were still there.
I then repeated the fsck on my real raid with the same command and then immediately ran a full backup on it.
So I am glad, I finally have my files back.
Regards, didiw