need help with software RAID

Subject: need help with software RAID
From: Steven Russell (
Date: Tue Feb 13 2001 - 21:53:24 MST

I posted the following to comp.os.linux.powerpc, but figured sending
it here, as well, would cover all my bases. BTW, the reformat
mentioned at the end did solve the error I was getting, but I still
can't boot from /dev/md0....

Thanks for any help.

-- Steven


Hi --

After failing to get software RAID to work with LinuxPPC installed, I
decided to give YDL (Champion Server 1.2.1) a go.... I'm trying to get
my system running using RAID-1 mirroring on all Linux partitions and was
hoping for some help...I did some reading of the software RAID HOWTO,
and have only gotten part way there.

First, my hardware:
   PowerPC 7200/120 64Meg of RAM
   2 internal 9 GB SCSI drives (/dev/sda, /dev/sdb)
   2 external ~2 GB SCSI drives (/dev/sdc, /dev/sdd)

/dev/sda was formatted such that:

   sda6 = 8+ GB (to become root partition)
   sda7 = 120MB (HFS standard format for MacOS)
   sda8 = 128MB (to become swap partition)

I formatted /dev/sdb similarly so that sdb6 and sdb8 can be mirrors of
their corresponding /dev/sda partitions. I'm planning on making sdc and
sdd into a single 2GB RAID allotment for /var (on the recommendation of
a friend). For now, however, sdc6 will be my root directory and sdd6
will be my swap.

- I installed Linux on /dev/sdc6 and it works fine.
- I created a RAID-level 1 device (/dev/md0) which syncronizes fine
using mkraid. (I had previously mounted /dev/sda6 at /mnt/sda6, and it
was also previously a formatted ext2 partition.)
- I changed /etc/fstab so that there's a line like:
      /dev/md0 /mnt/md0 ext2 defaults 1 2
- Reboot
- At startup, I get an error that the disk check (fsck) for md0 failed.
The message is something like:
     "/dev/md0: The filesystem size (according to the superblock) is
8704000 blocks
       The physical size of the device is 8703936 blocks
       Either the superblock or the partition is likely to be corrupt!"
- I either have to log in as root or reboot.
- After log in, if I type "mount /mnt/md0", the filesystem seems to
mount fine with no error.
- If I change the last two columns of the /etc/fstab entry to "0 0", the
disk check is skipped and I can mount the partition fine after login.

Now, it seems like the reported physical size of the disk is actually
the same block as where the superblock is, i.e. the superblock is, as I
understand it, at block 8703936. Is there something weird about my
filesystem selection, ext2?

Ultimately, I'd like to be able to boot off of one of the RAID
partitions. What steps do I need to take from here (a partially working
RAID system) to getting a fully redundant system up and running?

My plan is to get /dev/md0 working, use the "cd / ; find . -xdev | cpio
-pm /mnt/md0" approach from the "software RAID howto" and then set BootX
to boot off of /dev/md0 (after adjusting /etc/fstab). Then I'll rework
/var to mount off of /dev/md1 (sdc6/sdd6). How do I do this? Do I need
to mess around with recompiling the kernel given that RAID-1 support
seems to be going fine, currently, without having done so?

I'd be more than happy to furnish more details. For now, I think I'll
try reformatting /dev/md0 to see if that clears up the problem.

Thanks for any help.

-- Steven

This archive was generated by hypermail 2a24 : Tue Feb 13 2001 - 21:54:52 MST