Software RAID problem.

Ben Russo ben at muppethouse.com
Tue Dec 14 15:38:49 UTC 2004


I had a box with a RAID-0 device /dev/md4
it was defined in /etc/raidtab as:

	raiddev /dev/md4
	raidlevel 0
	nr-raid-disks	2
	chunk-size 	64k
	persistent-superblock	1
	nr-spare-disks	0
		device	/dev/hda7
		raid-disk	0
		device	/dev/hdc7
		raid-disk	1

All worked well for a year then one of the
disks died.  No problem I shutdown the system
replaced the failed drive with a new disk, and
then rebooted in single user mode,

	made a backup copy of the /etc/raidtab
	made a backup copy of the /etc/fstab
	removed the lines from raidtab/fstab
	    that refered to /dev/md4

Then I created the partition table on /dev/hdc
with fdisk (since all 3 disks are identical models
I used the EXACT same partition table from /dev/hda)

I ran badblocks on /dev/hda7 and /dev/hdc7 to make
sure all was good (this took a whole day).

Then I recreated /dev/md4 with the command:

	mdadm --create --verbose /dev/md4 --level=0 --raid-devices=2 /dev/hda7 
/dev/hdc7

After which I did a mkfs on /dev/md4 and then mounted it and
restored my data from backup.

Then I restored the backup copies of /etc/fstab and /etc/raidtab.
Here comes the problem.... When I rebooted I get dumped to single
user mode because the system detects a raid configuration error.
I log into single user mode with the root password, and try
	mount -a
I get:
	EXT3-fs error (device ide0(3,7)): ext3_check_descriptors: Block bitmap 
for group 1920 not in group (block 0)!
	mount: wrong fs type, bad option, bad superblock on /dev/hda7, or too 
many mounted filesystems

I know that I don't have too many mounted filesystems, and I am sure 
that the superblock on /dev/hda7 is fine, I mean I did a full
	"badblocks -fwsv -c4096 /dev/hda7"
just a few days ago and it was all OK.  Also, I was able to create the 
filesystem and restore all the data without a problem.

Now here is the issue.  I have over 300GB of data on this beast. I could 
rebuild the wholething AGAIN, but it will take a few more days to do it.

Does anybody see the problem?  What can I do to fix it?

Thanks in advance,
-Ben.




More information about the redhat-list mailing list