VG lost, no lvmconf => Help !

Here's my startup config :
on a single SCSI disk :
- linux 2.4.17 was booting on an lvm partiton, VG main, with LV : root, usr, boot, and var.
- an old linux system, 2.4.17 also, but classical ext2-based.

The LVM linux system was created by copying the root partitions of the ext2 system (as described in the LVM doc). Everything was working great with the LVM-based system, but i kept the old ext2 system as a rescue linux. Then, I did a vgextend to a remaining PV to my VG, and it boot and works already.

Problems arrived when i tried to add a SCSI CD burner : lvm didn't find the PVs, nor the VG, causing a kernel panic at boot time, and so did the ext2 linux. I was unable to fix that problem (SCSI id checked, ...), i removed that SCSI device, and then only the ext2 system did boot again. My vg was no longer detected.

So, here's where i am : under the ext2 system, i have no lvmtab for my VG, neither a recent main.conf, and my vg is no more detected, so i cannot recover my vg with vgcfgrestore. I did scan the 2 LVM PVs, and they are ok. So my question is :
is there a way to recover my VG with the datas contained in the PVs ?

info :

The system is a Linux-2.4.17 based kernel, with lvm-1.0.1-1

Here's my log :

////////------pvscan-------

# pvscan
pvscan -- reading all physical volumes (this may take a while...)
pvscan -- inactive PV "/dev/sda1"  is associated to an unknown VG (run vgscan)
pvscan -- inactive PV "/dev/sda3"  is associated to an unknown VG (run vgscan)
pvscan -- total: 2 [5.39 GB] / in use: 2 [5.39 GB] / in no VG: 0 [0]

///////------lvmdiskscan-------
-]# lvmdiskscan
lvmdiskscan -- reading all disks / partitions (this may take a while...)

lvmdiskscan -- /dev/sda1 [       3.99 GB] Primary  LVM partition [0x8E]
lvmdiskscan -- /dev/sda2 [       2.60 GB] Primary  LINUX native partition [0x83]lvmdiskscan -- /dev/sda3 [       1.39 GB] Primary  LVM partition [0x8E]
lvmdiskscan -- /dev/sda4 [     494.19 MB] Primary  LINUX swap partition [0x82]

lvmdiskscan -- 1 disk
lvmdiskscan -- 0 whole disks
lvmdiskscan -- 0 loop devices
lvmdiskscan -- 0 multiple devices
lvmdiskscan -- 0 network block devices
lvmdiskscan -- 4 partitions
lvmdiskscan -- 2 LVM physical volume partitions

///////------vgscan--------
]# vgscan
vgscan -- reading all physical volumes (this may take a while...) vgscan -- "/etc/lvmtab" and "/etc/lvmtab.d" successfully created vgscan -- WARNING: This program does not do a VGDA backup of your volume group

///////------vgdisplay-----
]# vgdisplay
vgdisplay -- no volume groups found

///////------pvdata-------
]# pvdata -v -V -L /dev/sda1 | grep -v empty
--- Volume group ---
VG Name

VG Access             read/write
VG Status             NOT available/resizable
VG #                  0
MAX LV                256
Cur LV                4
Open LV               0
MAX LV Size           255.99 GB
Max PV                256
Cur PV                2
Act PV                2
VG Size               5.38 GB
PE Size               4.00 MB
Total PE              1376
Alloc PE / Size       1277 / 4.99 GB
Free  PE / Size       99 / 396.00 MB
VG UUID               ML3C0g-bpt6-2He5-AK6f-5DNt-2LVV-bEm05U

--- List of logical volumes ---

--- Logical volume ---

LV Name                /dev/main/boot
VG Name                main
LV Write Access        read/write
LV Status              NOT available
LV #                   1
# open                 0
LV Size                40.00 MB
Current LE             10
Allocated LE           10
Allocation             next free
Read ahead sectors     120
Block device           58:0

read_ahead: 120

--- Logical volume ---

LV Name                /dev/main/root
VG Name                main
LV Write Access        read/write
LV Status              NOT available
LV #                   2
# open                 0
LV Size                1.00 GB
Current LE             256
Allocated LE           256
Allocation             next free
Read ahead sectors     120
Block device           58:1

read_ahead: 120

--- Logical volume ---

LV Name                /dev/main/usr
VG Name                main
LV Write Access        read/write
LV Status              NOT available
LV #                   3
# open                 0
LV Size                3.00 GB
Current LE             768
Allocated LE           768
Allocation             next free
Read ahead sectors     120
Block device           58:2

read_ahead: 120

--- Logical volume ---

LV Name                /dev/main/var
VG Name                main
LV Write Access        read/write
LV Status              NOT available
LV #                   4
# open                 0
LV Size                972.00 MB
Current LE             243
Allocated LE           243
Allocation             next free
Read ahead sectors     120
Block device           58:3

read_ahead: 120

pvdata -- logical volume struct at offset 4 is empty
....

and i have the same for /dev/sda3

///////////----------------

So, how can i use this stuff to recreate a good VG info and make my system back ? Many thanks,

Guillaume

Can you send "vgscan -d" to me in personal email ([email protected])? What does lvm-1.0.1-1 mean. We don't have such version. Please check with for eg. "vgdisplay -h".

<SNIP>

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
[email protected]                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

OK, i'll do that.

In fact, 1.0.1-1-mdk (Mandrake), you can find it on rpmfind.net

---------------
vgdisplay -h
Logical Volume Manager 1.0.1
Heinz Mauelshagen, Sistina Software 26/11/2001 (IOP 10)

vgdisplay -- display volume group information

vgdisplay [-c|--colon | -s|--short | -v[v]|--verbose [--verbose]]

        [-d|--debug] [-h|--help] [--version]
        [-A|--activevolumegroups | [-D|--disk] [VolumeGroupName...] ]

---------------

A little more info :

i tried to reimport the old PVs
----------------
]# vgimport -f -v main /dev/sda1 /dev/sda3
vgimport -- locking logical volume manager
vgimport -- checking volume group name
vgimport -- checking volume group "main" existence
vgimport -- trying to read physical volumes
vgimport -- checking for duplicate physical volumes
vgimport -- checking physical volume name "/dev/sda1"
vgimport -- reading data of physical volume "/dev/sda1" from disk vgimport -- checking for exported physical volume "/dev/sda1"
vgimport -- reallocating memory
vgimport -- checking consistency of physical volume "/dev/sda1"
vgimport -- checking for duplicate physical volumes
vgimport -- checking physical volume name "/dev/sda3"
vgimport -- reading data of physical volume "/dev/sda3" from disk vgimport -- checking for exported physical volume "/dev/sda3"
vgimport -- reallocating memory
vgimport -- checking consistency of physical volume "/dev/sda3"
vgimport -- physical volumes "/dev/sda3" and "/dev/sda1" are in different volume groups

----------------

Next, i said my pvdata was identical for /dev/sda1 and /dev/sda3, but i discover that was'nt true :
---------------
#]pvdata -P /dev/sda3
--- Physical volume ---

PV Name               /dev/sda3
VG Name               main
PV Size               1.39 GB / NOT usable 7.65 MB [LVM: 125.00 KB]
PV#                   2
PV Status             NOT available
Allocatable           yes
Cur LV                1
PE Size (KByte)       4096
Total PE              355
Free PE               99
Allocated PE          256
PV UUID               none

----------------
see that PVUUID ?

So i did try uuid_fixer, which finally give me a human readable error :
----------------
]# ./uuid_fixer /dev/sda1 /dev/sda3
/dev/sda3 - UUID corrupt
----------------
Damned.
So where do we go now ?
thanks,

Guillaume

uuid_fixer aims to write a correct list of PV uuids to every PV. But it can't handle this case, where you don't have a PV UUID at all on /dev/sda3.

So, let's hack one in to hopefully make LVM happier.

First preserve the metadata on sda1 and sda3 for recovery with:

dd if=/dev/sda1 of=sda1.VGDA count=512 bs=1k
dd if=/dev/sda3 of=sda3.VGDA count=512 bs=1k

Backup sda[13].VGDA please.

echo "sahdgdgasjg21hxjn45jxJJ45aasdc3d"|dd of=/dev/sda3 obs=1 seek=44 count=32

Run

./uuid_fixer /dev/sda1 /dev/sda3

again.
Retry vgscan afterwards.

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
[email protected]                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

done ;-)

on a snapshoted nfs.

I didn't remember enough C to calculate this offset ;-) ! Question : is it a pure random UUID, that uuid_fixer will propagate wherever it is needed ?

uuid_fixer found the 2 PVs with their respective UUID,
when asked about writing in each file, i didn't find what it was about in the README, but i answered affirmative, and then normal termination. But vgscan missed again my VG.

Some few uuid_fixer and failing scans later, i tried "no" to the mistery question of uuid_fixer, after what vgscan find up my "main" VG.

vgchange -ay, mounting, cfgbackup, lilo, etc., everything was working like a charm ! Either did the reboot, so i'm currently running with my lvm root partition. Many thanks, many thanks, everything is fixed.

So I have only some informative question : what did uuid_fixer do, and what does the question mean ?

Guillaume

:)

Ok

Well, to tell you the truth: it was my fingers running over my keyboard in a nervous coffein lack ;-)
And yes, uuid_fixer should.

Hmmm. Strange.
If you answer yes or use the force option it should have written the PV UUID lists to the devices correctly.

Like: "uuid_fixer -f /dev/sda[13]"

You seem to have disks which need to be convinced to take writes ;-)

You're welcome.

As said above:
the question comes up unless you use uuid_fixer with the "-f" option. If you answer 'y', it writes the PV UUID list back to all the devices on the command line. PV UUID lists of *all* PVs are stored on all PVs of a VG in order to identify the relationship between VG and PVs uniquely.

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
[email protected]                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-