Hi,
I accidently hosed my RAID1 and 5 systems by overwriting the first 512 bytes of each physical disk participating in the RAID. Needless to say, I am past kicking myself in the pants, and am trying to retrieve my data and get my RAID1 and RAID5 disks functional again. My setup was four 160GB IDE disks, indentical in size and partitioned identically (1024MB 0xfd, 256MB 0x82, rest of drive being 0xfd). The 1st partition was setup as RAID1 with 2 mirrors and 2 spares, and the 3rd partition was setup as RAID5.
I noticed my array was broken when fsck tried checking my LVM disk, which had previously failed mount because the raid driver couldn't find RAID partitions. So I regenerated the partition tables to (what I thought) were the actual sizes and locations of the partitions on the disk previously. I recall choosing "nice" easy-to-remember numbers for the sizes of the physical partition (like 1024MB, 256MB, "rest of disk"). I use standard linux fdisk to partition drives.
I then edited my /etc/raidtab and changed all the "raid-disk" directives to "failed-disk" and ran "mkraid --force /dev/md0" to see what It would do. Lo and behold, it failed to find the md superblock on each one of the four newly created partitions (one on each disk). This surprised me, because to my knowledge, nothing else on the disk had been overwritten:
# mkraid --really-force /dev/md0
DESTROYING the contents of /dev/md0 in 5 seconds, Ctrl-C if unsure!
handling MD device /dev/md0
analyzing super-block
disk 0: /dev/hde1, failed
disk 1: /dev/hdi1, failed
disk 2: /dev/hdg1, 102384kB, raid superblock at 102272kB
disk 3: /dev/hdk1, 102384kB, raid superblock at 102272kB
mkraid: aborted.
(In addition to the above messages, see the syslog and /proc/mdstat as well
for potential clues.)
#
(afaict, you can ignore the 102384kB raid superblock markers...those were created in my first attempt at making a raid a couple months ago...I didnt bother zeroing the disks before remaking my raid partitions). I am actually quite surprised that my ancient raid superblock from a 100MB partition coincidently ended up exactly where mkraid was looking for the raid superblock for my 1024MB partition!)
Though gpart and rescuept don't look for RAID partitions, I attempted using them anyway and found traces of my partitions all over the disks (I used ext3 on top of LVM on top of the RAID5). I read linux-2.6.9/drivers/md/md.c and linux-2.6.9/include/linux/raid/md_p.h and modified rescuept to look for RAID superblocks. I found them, and they were completely intact, and located in the "general" area of where I expected them to be (i.e. near the end of each physical partition). However, my attempts at trying to determine the exact location of the physical partition from the location of the raid superblock and the size of the "enclosed partition" (sb->size) has not given me much luck (I probably am making erroneous assumptions about where exactly the raid superblocks are located).
Is there any kind of software setting or tool that those in-the-know can recommend that I use for determining the locations of the physical partitions based on the location of the RAID superblock? (or anything else for that matter!)
The information on my RAID5 array is much more important to me than the data on my RAID1 array (used only to boot the RAID), so I have not attempted to use mkraid --force on /dev/md1 until I successfully retrieve data from /dev/md0. (It is not clear to me whether mkraid --force writes anything to the physical disk or not).
It doesn't feel like the time to throw in the towel and repartition/recreate my RAID, so any and all help and advice is greatly appreciated!
Thanks,
- Armen
- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
