Skip to contentSkip to conversation

Newbie: Yet another "Incorrect metadata area header checksum"

linux-lvm Simon Waters 3 participants 5 messages Feb 1, 2010 – Feb 2, 2010

lore
Share & tools
lore
#0 lore raw

Installed Debian Lenny using NetInst CD 5.0.3 i386 dated 20090906-12:06 onto a DELL SC1425 with two SATA disks.

The system had previously had Centos 5.4 installed with LVM. The Debian installer was fighting back a little (because of this?), but eventually I picked "Guided Install with LVM" and it behaved in a plausible fashion (/dev/sda1 for /boot, /dev/sda2 for LVM, lots of partitions because I asked for them).

All seemed to go fine. The system installed, and booted, I shrunk /home file system and then logical volume, then expanded /home to fit the new lv size. I add /dev/sdb1 to the only volume group, add a logical volume. Basically convinced myself LVM was working and that I could administer it.

The I turned my attention to the error message that some of the commands
produced: "Incorrect metadata area header checksum", since it looked annoying.

I've tried a lot of things rather blindly. The first occurance is in the installer log after the message "Queuing package e2fsprogs for later installation" from "Main-Menu" so presumably in response to me using the Debian installer menu's to set up LVM. This suggests it all went pear shaped before I had the opportunity to mistype stuff at the command line.

One suggestion was to install "dmraid" and look for metadata using this. It produces:
/dev/sdb: ddf1. ".ddf1_disks", GROUP, ok, 155989151 sectors, data@ 0 /dev/sda: ddf1. ".ddf1_disks", GROUP, ok, 155989151 sectors, data@ 0

Attempts to erase this with "dmraid -rE" as suggested in one post fail with a three lines of:

ERROR: seeking device "/dev/sdb" to 40959999737856
ERROR: writing metadata to /dev/sdb, offset 79999999488 sectors, size 0 bytes returned 0
ERROR: erasing ondisk metadata on /dev/sdb
(and same for /dev/sda).

Other suggestion was make sure Debian Lenny packages were upto date. They are. The netinst CD just does that now - I'm sure it didn't always do that but I double checked and no updates are available.

This is my first (Linux) LVM install - be gentle.

Did I fail in preparing my disks in some way?

#1 lore raw
Quoted 2 lines · Simon Waters

Installed Debian Lenny using NetInst CD 5.0.3 i386 dated 20090906-12:06 onto a DELL SC1425 with two SATA disks.

....

Quoted 5 lines

One suggestion was to install "dmraid" and look for metadata using this. It produces:
/dev/sdb: ddf1. ".ddf1_disks", GROUP, ok, 155989151 sectors, data@ 0 /dev/sda: ddf1. ".ddf1_disks", GROUP, ok, 155989151 sectors, data@ 0

i believe your dell machine configured the two disks in a fakeraid, yet Lenny does not know (you did not tell it or you forced it to ignore the fakeraid) and tries to use the two disks independently. This will lead to problems since at every boot the fakeraid bios will touch the two disks.

back up your data
decide wether you want to use the fakeraid or not
configure your bios accordingly
reinstall your OS accordingly

L.

Signature · 6 lines
Luca Berra -- [email protected]
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \
#2 lore raw
Quoted 7 lines · On Tuesday 02 February 2010 06:55:48 Luca Berra wrote:
Quoted 5 lines

One suggestion was to install "dmraid" and look for metadata using this. It produces:
/dev/sdb: ddf1. ".ddf1_disks", GROUP, ok, 155989151 sectors, data@ 0 /dev/sda: ddf1. ".ddf1_disks", GROUP, ok, 155989151 sectors, data@ 0

back up your data

I haven't restored it to this machine yet - so that bit was easy.

Quoted 2 lines

decide wether you want to use the fakeraid or not
configure your bios accordingly

I'd decided not already after appalling experiences of fakeraid from DELL before.

The BIOS has the embedded SATA controller set to "ATA Mode" not RAID, and this was the same throughout the installation.

Setting the BIOS to RAID revealed one RAID-1 array, which one can then delete, and then converting the system back to ATAMODE, and everything on disk is well and truely hosed (hence the backup suggestion no doubt, and several dire warnings from the BIOS). I looked to see if I could recover anything from this, as I thought it might be useful exercise in recovering systems but it didn't look hopeful.

Perils of state buried in an inherited machine. I note that the RAID BIOS reports that the state of the RAID-1 array was fine, I think I'm remembering why I've come to hate the fakeraid stuff. I'm also not sure about the wisdom of confirming pressing the "Y" key by asking the user to press the "Y" key again.

Anyway I've now reinstalled after the above, and "dmraid -r" now reports "No RAID disks".

However in the installer log I still have the same error about incorrect metadata area header checksum.

The is first reported looking for the old Volume Group (so the installer found my old LVM information, even if I couldn't with the installer in rescue mode). But when partman-lvm in the Debian installer starts it reports.

Volume group "h78" successfully created

/dev/cdrom: open failed: Read-only file system

Incorrect metadata area header checksum

Only thing I did outside Debian Lenny Guided install with LVM was to recreate "root" bigger and "home" smaller.

Excluding the /dev/cdrom device in /etc/lvm/lvm.conf doesn't help.

Is this due to old metadata on the disk, or is there something going wrong in the installation process.

Since /dev/sda2 contains the root volume (amongst others) I assume any attempt to reset this data will require me to boot from a rescue CD?

Simon

#3 lore raw
Quoted 2 lines · On Tuesday 02 February 2010 10:21:30 Simon Waters wrote:

Is this due to old metadata on the disk, or is there something going wrong in the installation process.

dd if=/dev/zero of=/dev/sda
dd if=/dev/zero of=/dev/sdb

From the Debian install CD, and then a reboot (otherwise the Debian installer remembers the old LVM information it read from disk when it booted, and complains things aren't like the use to be.

#4 lore raw
Quoted 10 lines · Simon Waters
Quoted 2 lines · On Tuesday 02 February 2010 10:21:30 Simon Waters wrote:

Is this due to old metadata on the disk, or is there something going wrong in the installation process.

dd if=/dev/zero of=/dev/sda
dd if=/dev/zero of=/dev/sdb

Quoted 1 line

From the Debian install CD, and then a reboot (otherwise the Debian installer

remembers the old LVM information it read from disk when it booted, and complains things aren't like the use to be.

Above wipes entire disk - which is good for a fresh install. But just making sure. Not so good if you wanted to keep some partitions...

A faster way to wipe the entire disk is to use the security erase feature of SMART disks available via /sbin/hdparm.

Signature · 4 lines
	      Stuart D. Gathman <[email protected]>
    Business Management Systems Inc.  Phone: 703 591-0911 Fax: 703 591-6154
"Confutatis maledictis, flammis acribus addictis" - background song for
a Microsoft sponsored "Where do you want to go from here?" commercial.