Bug#548648: misdetects md RAID UUID

Ian MacDonald imac at netstatz.com
Tue Aug 31 05:25:45 UTC 2010


I have seen this problem when migrating systems from a single ext[234]
disk to RAID1. The problem seems to be that when an ext filesystem is
overwritten with an md device, the UUID of the original ext file system
is preserved, or continues to be detected.

In my cases the md device represents the entire /boot partition, so it
always contains a valid ext filesystem (each disk can be mounted as
individual ext in a rescue scenario even though they are part of a
raid1). It seems in the RAID1 case, the inner ext UUID should be ignored
based on the partition type which is set to raid_member.

The interesting part is that this ONLY happens when the disk was an ext
partition PRIOR to being mdraid.  In my case sdb1 was the original
ext2 /boot partition, then md0 was started on sda1 with 1 of 2 disks
active, the data was migrated from /dev/sdb1 to /dev/md0, then /dev/sdb1
was repartitioned and added to the /dev/md0 array.

I offer some additional output, as it is blkid that is detecting the
disk correctly (per mdadm), and grub and tune2fs are showing the wrong
UUID. I removed some unrelated info on other devices from the blkid and
mdstat output, but the problem is clear and easily repeatable using the
process described above. 

nexus:/boot# tune2fs -l /dev/md0 | grep UUID
Filesystem UUID:          4075dce0-fb12-4017-bf1d-1b2d86eee0ec

nexus:/boot# grub-probe --device /dev/md0 --target=fs_uuid
4075dce0-fb12-4017-bf1d-1b2d86eee0ec

nexus:/boot# mdadm -D /dev/md0 | grep -i uuid
           UUID : aeaeb946:1e6654da:ca9135eb:329f23c1 (local to host nexus)

nexus:/boot# blkid 
/dev/sdb1: UUID="4075dce0-fb12-4017-bf1d-1b2d86eee0ec" TYPE="ext2" 
/dev/sda1: UUID="aeaeb946-1e66-54da-ca91-35eb329f23c1" TYPE="linux_raid_member" 
/dev/md0: UUID="4075dce0-fb12-4017-bf1d-1b2d86eee0ec" TYPE="ext2" 

nexus:/boot# cat /proc/mdstat 
Personalities : [raid1] 
md0 : active raid1 sda1[0] sdb1[1]
      248896 blocks [2/2] [UU]
      
This is currently causing me some pain on two servers .. where I have
just given up on RAID1 /boot for the time being which is less than
ideal.






More information about the Pkg-grub-devel mailing list