RAID 1 error question - boot problem.

[Bill Davidsen]

Robin Laing wrote:
> Hello,
> 
> I am trying to help someone get a system to reboot after a system issue. 
>    He is not the builder of the system and the person that knows the 
> system is away for a few weeks.  Great timing.  :)
> 
> It is Fedora 10.
> 
> Two drives, two partitions each drive which one is a mirror.
> 
> /dev/sda is partitioned as /boot and / which is mirror 1.
> /dev/sdb is partitioned as swap and /
> 
> The system wouldn't restart on the reboot and came up with an error 
> after creating the raid arrays and then saying that it cannot find 
> /dev/md0.  I don't have the exact error message right now.
> 
> Using an Ubuntu disk (persons personal preference) the system was booted 
> into a live system and using gparted the partitions were shown to be as 
> above with as I see it, one error.
> 
> /dev/sda1 ext3 boot
> /dev/sda2 ext3 /   raid
> 
> /dev/sdb1 Swap
> /dev/sdb2 Unknown  / raid
> 
> Running mdadm /dev/sdb2 --examine shows that the partition superblock is 
> showing RAID 1 and that it is clean.
> 
> As this is a critical system, it is a priority and is being used as a 
> virtual server.
> 
> With only the second drive installed, we tried to run fsck.ext3 on the 
> /dev/sda2 (normally b2) with no success.  We also tried /dev/md0 as 
> Ubuntu has created the /dev/md0 from the single drive.
> 
> The user has not tried to boot with only the one drive in yet.  He is 
> making a copy of the drive on a different system.
> 
> Now, the question.  On booting from a mirror 1 array, if there is a 
> problem with the raid system, how does the boot process read the 
> mdadm.conf file when it is on the RAID array that needs to be created? 
> Is there some data that is stored in the /boot or someplace else that 
> has the necessary info to tell the system how to build the array?
> 
> Is it part of the /boot/grub/device.map or /boot/System.map* ?
> 
> Any suggestions to where to start?
> 
The linux-raid group would have been a better choice, but this is a simple 
question. The mdadm.conf file should get put in the initrd file, which is in the 
/boot partition, which you didn't mirror for some reason. I'm guessing that sda2 
is a better place to start, since that's recognizable as an ext3 partition. 
Having a partition identify as "Unknown" is usually not a good thing. I would 
mount that partition and copy the contents to a secure backup if this is critical.

"I don't have the exact error message right now" doesn't help, I suggest backing 
up sda2, and sda1 if you can, noting the error message, and post back. Without 
more information I am guessing that the sdb2 partition is in some way hosed, do 
NOT run fsck on sda2 before backing up, and run with the "-n" option to see what 
condition the f/s is in. I doubt you've lost your data yet, don't do anything 
which would change that.

-- 
Bill Davidsen <davidsen at tmr.com>
   "We have more to fear from the bungling of the incompetent than from
the machinations of the wicked."  - from Slashdot