All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Andre Bonhote" <andre@bonhote.com>
To: Linux-LVM@Sistina.com, Mauelshagen@sistina.com
Cc: beat@rubis.ch
Subject: [linux-lvm] LVM Recovery (urgent)
Date: Wed Jan  8 08:34:02 2003	[thread overview]
Message-ID: <EHEDIKGDCNFCNANFPGCDAENOCAAA.andre@bonhote.com> (raw)

Hi Heinz

Since I was not sure if you really understand the german tongue, so I write
in english. You're really kind of my last resort concerning my problem.

You might ask why I don't write to one of the mailing lists. Well, the
answer is quite simple: I don't trust them in this case. I don't think I
have many tries left to save my data, so I need a real expert advice. Don't
ask about a backup, there's none. And this is just one of my major faults.
But maybe I found a bug, we'll see. Let me describe my problem and what I
did first.

My setup is as following:
 - 2 SCSI drives, 18/9 GB
 - 1 IDE drive, 80 GB
 - Redhat 8.0, with
   - / on LVM (!! Redhat allows this without warning. I didn't know about
the
     dangerosity)
   - /boot on a standard ext3 partition
   - /home, respectively /new-home on LVM, which is my major concern.
 - I have/had one VG, /dev/CaradhrasVG/ spanning first the two SCSI disks.
   Yesterday I added the IDE disk to this VG.
   - /dev/CaradhrasVG/RootVG holding the /(remember this name!). Here, also
     /new-home and /home is located
   - /dev/CaradhrasVG/HomeLV, formatted with ext3 but empty. It was intended
to
     copy the homes here lateron.

Fdisk shows this:
/dev/sda:
 255 heads, 63 sectors, 1106 cyls
 /dev/sda1,Bootable,Start:1,End:13,Blocks:104391,83/Linux
 /dev/sda2,-,Start:14,End:1106,Blocks:8779522+,8e/LinuxLVM

/dev/sdb
 2231 cyls
 /dev/sdb1,Bootable,Start:1,End:2231,Blocks:17920476,8e/LinuxLVM

/dev/hdc
 fdisk -l shows no output. There's no partitioning. I did pvcreate /dev/hdc.

Ok, yesternight I added my IDE drive /dev/hdc to the VG. With the new space
inside the VG I decided to create the HomeLV using the whole new space.
Doing this (which ran fine and without any hassle), I found that the Root
Volume had the wrong name (RootVG instead of RootLV). So I changed its name
using "lvrename". After that, I changed the entry in /etc/fstab. First big
mistake. The result was, that no booting was possible with an error message
like that:

 mount: error 2 mounting ext3
 pivotroot: pivot_root(/sysroot,/sysroot/initrd) failed: 2
 umount /initrd/proc failed: 2

A search on google brought exactly two japanese pages. A different search
showed up a lot of different error codes, but not "error 2".

Since Redhat offers the "linux rescue" mode on their CD, I tried that, and
voil�! it worked. But here, I obviously did my second (but less serious
IMHO) mistake. Without being in a chrooted environment, I changed the name
of RootLV back to RootVG. After booting from disk, my RH installation booted
and found init but failed to fsck /. Simply because the Volume was still
somehow named RootVG, but mounted as RootLV.

Then all went very fast. It was almost three o'clock in the morning and I
was very tired and angry about myself, this is obviously one of the reasons
why I commited such stupid, stupid mistakes. But here we go.

I decided to extract the (unused? no!) IDE drive from the VG to store the
important data in some safe place. Again, with RH's rescue CD but WITHOUT
chroot. When I issued "vgreduce CaradhrasVG /dev/hdc", vgreduce answered
with an error message like "it's in use. don't do that". Well, I thought in
my fever that it means the VG which was in use. Therefore I rebooted with
the CD but I didn't mount the OS on the disks at all, which I thought would
allow me to vgreduce. It worked, but with an error message which I don't
remember. I personally think that the label on /dev/hdc has been removed but
no (which?) config files have been updated. Could this be a bug? Well,
anyway, after a reboot with the RH CD there was just the message "Could not
find a valid RedHat Installation". Bang-boom. "vgscan -v" shows this:

...
reading data of volume group "CaradhrasVG" from physical volume(s)
ERROR "vg_read_with_pv_and_lv(): current PV" can't get data of volume group
 "CaradhrasVG" from physical volumes
...

For me, this shows that if /dev/hdc would be added to the VG again, it could
work.

You know, I don't care about my personal data. But on these disks, we have
around 3100 pictures of our kids of which I don't have a solid backup (some
months are missing), and all my wife's E-Mails (which was just too big to
store on a CD Rom) plus some very, very old compuserve mails which document
how my wife and I met several years ago. It's memory, without backup. That's
also why I don't trust the mailing lists: It's just too important to me and
my wife! I hope you understand that and accept it as a reason why I am
bothering you personally.

I don't know the LVM internals, but I have several ideas how I could bring
the data back. I am kindly asking you to point out IF there's a way to fix
it, and if yes, how I could do it. Here's what I was thinking of.

- I could try (under your lead if possible) to bring /dev/hdc back into the
CaradhrasVG
  so that I can redo what I wanted to (extract hdc), but CLEANLY this time.
IMHO this is
  the safest solution, if it's possible. As I said, I don't know how LVM
rally works,
  but I guess that you store information about the VG membership somewhere
on the drive
  itself. If yes, there could be a chance to recover that (hopefully).

- The second approach (which I don't like that much) is to remove the hdc
from the VG
  somehow. But since I don't have access to /etc, I have no clue how to
manage this.

- The third possible approach (which I really don't like) is to try to
access /home and
  /new-home on the VG directly somehow. I am quite sure that only the SCSI
disks are used
  by the VG, but not 100%

This mail will go in CC to Beat Rubischon, a veteran unix admin and head of
the LUGS. He agreed to help me sorting out this mess.

Heinz, I really hope you feel my desesperation. I have no clue how to
proceed best from here, but I also don't want to accept that all data is
lost. It's just too emotionally important to us. Memories.

Thank you very much in advance. I am really looking forward to hearing from
you soon, and if it's only a line like "yes, I have read your mail and I am
thinking about it" or something similar. This would set me a bit at ease.

Yours Sincerely

Andr� Bonhote

--
Andre Bonhote - Foehrenweg 6 - CH-8956 Killwangen
P: +41564011734  M: +41797069322  F: +41564011712
Pers: andre@bonhote.com Bus:andre.bonhote@colt.ch

             reply	other threads:[~2003-01-08  8:34 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2003-01-08  8:34 Andre Bonhote [this message]
2003-01-09  3:49 ` [linux-lvm] Re: LVM Recovery (urgent) Andre Bonhote
2003-01-10  1:35 ` [linux-lvm] Re: LVM Recovery - mostly solved Andre Bonhote

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=EHEDIKGDCNFCNANFPGCDAENOCAAA.andre@bonhote.com \
    --to=andre@bonhote.com \
    --cc=Linux-LVM@Sistina.com \
    --cc=Mauelshagen@sistina.com \
    --cc=beat@rubis.ch \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.