lvlnboot: cannot query phys.vol. ""

[Ceesjan van Hattum]

I thought to have a perfect cloning procedure. But now an a new system, I rebuild the lvmtab (by vgscan -v) and try to rebuild the vg00.conf by lvlnboot -R.
I have seen the same error 'with' a disk in case i forgot to 'vgchange -a y /dev/..'.

How can i resolve this ""-disk without rebooting the system?

[Paula J Frazer-Campbell]

HI

Try moving lvmtab to lvmtab.bak and rebuilding it.

vgscan -a


Paula

[Ceesjan van Hattum]

mv /etc/lvmtab /etc/lvmtab.old
vgscan -a
Looks alright, but..., but no result: same error on lvlnboot -R, vgdisplay etc.

..other suggestion?

[Paula J Frazer-Campbell]

Hi

Have a look in /dev at all of the VG dirs - do you have any strange entries there ?

If so cp/mv/rm them and do the lvmtab.bak and vgscan -a again.

Example below


Paula

root@K1/dev/vg00>ll
total 0
crw-r----- 1 root sys 64 0x000000 Jul 23 2001 group
brw-r----- 1 root sys 64 0x000001 Aug 27 10:48 lvol1
brw-r----- 1 root sys 64 0x000002 Jul 23 2001 lvol2
brw-r----- 1 root sys 64 0x000003 Jul 23 2001 lvol3
brw-r----- 1 root sys 64 0x000004 May 7 13:03 lvol4
brw-r----- 1 root sys 64 0x000005 May 7 13:03 lvol5
brw-r----- 1 root sys 64 0x000006 May 7 13:03 lvol6
brw-r----- 1 root sys 64 0x000007 Aug 2 15:43 lvol7
brw-r----- 1 root sys 64 0x000008 Aug 2 15:43 lvol8
crw-r----- 1 root sys 64 0x000001 Jul 23 2001 rlvol1
crw-r----- 1 root sys 64 0x000002 Jul 23 2001 rlvol2
crw-r----- 1 root sys 64 0x000003 Jul 23 2001 rlvol3
crw-r----- 1 root sys 64 0x000004 Jul 23 2001 rlvol4
crw-r----- 1 root sys 64 0x000005 Jul 23 2001 rlvol5
crw-r----- 1 root sys 64 0x000006 Jul 23 2001 rlvol6
crw-r----- 1 root sys 64 0x000007 Jul 23 2001 rlvol7
crw-r----- 1 root sys 64 0x000008 Jul 23 2001 rlvol8

[Ceesjan van Hattum]

Everything looks really normal. Only /etc/lvmconf/vg00.conf is still incorrect and needs to be rebuild, but cannot due this error by lvlnboot -R.
All other structures (incl. dir-structs) seem to be good, so i decided to reboot the system (very lucky this one is no production (yet)): no result. :(
Any other hints my friend?

[Paula J Frazer-Campbell]

Hi

What does strings /etc/lvmconf/vg00.conf say ? Does it have any strange bits ?


Example:-


root@K1/etc/lvmconf>strings vg00.con
CONFIG01
/dev/vg00
/dev/rdsk/c3t6d0
/dev/rdsk/c2t4d0
/dev/rdsk/c3t5d0
LVMREC010C
HPLVMBDR<
LVMREC010C
LVMREC010C
VGDA0001
VGSA0001=fK


Paula

[Paula J Frazer-Campbell]

Hi

Also have you tried vgcfgbackup?

Man vgcfgbackup deals mainly with vg00.

Paula

[Paula J Frazer-Campbell]

Sorry I forgot

Then use vgcfgrestore

to rebuild - this may clean the file on either the backup or restore - make a copy before trying this.

Paula

[Ceesjan van Hattum]

Hi,
$> strings vg00.conf
CONFIG01
/dev/vg00
/dev/rdsk/c1t0d0
/dev/rdsk/c0t3d0
/dev/rdsk/c0t0d0
LVMREC01
HPLVMBDR=l
LVMREC01
LVMREC01
HPLVMBDR=l
VGDA0001
VGSA0001=l

Looks quit normal to me. I tried to do
vgcfgbackup -f /tmp/vg00.backup vg00
but... on this command same error:
vgcfgbackup: Cannot query physical volume "".

The vg00.conf need to be rebuild (delete of c0t3d0), so i also tried:
mv /etc/lvmtab /etc/lvmtab.old
vgscan -a
vgimport -v /dev/vg00 /dev/dsk/c0t0d0 /dev/dsk/c1t0d0
vgchange -a y /dev/vg00
.. but no results..

Secondly i mv /dev/vg00 in the hope i could recreate it, but i was not able to.

One thought:
if i try lvlnboot -R, i got 11 times extra this error. Likely due to the fact of having 11 lvols on board.
This means to my understanding, it reads /dev/vg00/rlvolXX ?? Is that correct?

[Sajid_1]

hello,

from what I read, I would suggest you to do this on the *problematic* disk:
# diskinfo /dev/rdsk/cxtydz
# pvdisplay -v /dev/dsk/cxtydz

hth

[James R. Ferguson]

Hi Ceesjan:

I suggest that you look at Technical Knowledge Base document #KBRC00000827. It may offer some further insight.

Regards!

...JRF...

[Ceesjan van Hattum]

Hi, this time i have some warning that contains
a important clue (i think):
vgchange -a y /dev/vg00
vgchange: Warning: Couldn't attach to the volume group physical volume "".
A component of the path of the physical volume does not extist.
Volume "/dev/vg00" has been succesfully changed.

Hmmmm :(
Where does the vg keeps it info about PV's ?
A pvdisplay i can not do, because of path-argument is needed.
James: Where i can find this doc? Search on doc-id had no hits...
Maybe there is an other way to find this doc?

All help will be appreciated.
ps. I have a 12 hours deadline to solve this issue.. :((

[Ceesjan van Hattum]

Reading some more manuals, i only just now learned about /etc/lvmpvg... this file does not exist.
On an other system (11.00) it does exist but is empty. On the machines we are talking about (10.20) it is just not there...
what to do?

[James R. Ferguson]

Hi (again) Ceesjan:

Make sure that you are logged into the ITRC to enable your complete access rights to the Technical Knowledge Base. Choose search by DOCID in the dropdown. The document number is correct -- KBRC00000827.

'etc/lvmtab' lists the physical devices that are associated with each volume group. Use 'strings /etc/lvmtab' to look at the binary file.

A 'vgscan' recreates 'etc'lvmtab'. It appears that you did this. By renaming the current 'etc/lvmtab' and then doing a 'vgscan', all disks on the system are scanned and a new 'lvmtab' is built from scratch.

LVM deposits a VGID (volume group ID) at a fixed address on the disk when a 'vgcreate' is done. When a physical disk is added to an existing volume group, the VGID is replicated to the disk being added. In this fashion, 'vgscan' can determine "who" belongs to "what" volume group.

For the missing physical device, try:

# diskinfo -v /dev/rdsk/cXtYdZ
# dd if=/dev/rdsk/cXtYdZ of=/dev/null bs=10k count=30

If you do not see a correct size or there are errors from the 'dd', you have a hardware problem.

Also try:

# lvdisplay -v /dev/vgXX/lvol${N}

...where N represents every logical volume on the disk. Look for missing extents shown as "???".

Regards!

...JRF...

[James R. Ferguson]

Hi (again) Ceesjan:

'etc/lvmpvg' is an optional file used when you want to allocate mirrors or otherwise 'lvextend' to specific groups of disk. This file is commonly used with "extent-based mirror stripping". See 'man 4 lvmpvg'.

Regards!

...JRF...

[Paula J Frazer-Campbell]

Hi

What does a :-
strings /etc/lvmtab show?

Paula

[Ceesjan van Hattum]

Okay, i will not longer look at lvmpvg.

The /etc/lvmtab shows:
/dev/vg00
/dev/dsk/c0t0d0
/dev/dsk/c1t0d0
Nothing special. I also restored the /etc/lvmconf by moving and recovering from identical (org from clone) machine. No results.
I also deteled all /dev/dsk/* and recreated by insf -e.
I recreated /etc/lvmtab by vgscan, but still could not recreate /etc/lvmconf/vg00.conf by lvlnboot -R.
On all lvm-related commands i have the "disk-with-no-name" called "". Sometimes as warning, sometimes as error.

Due to lack of time, i decided to start from scratch. It might be that i made some mistake somewhere in the beginning of the procedure. I emptied all disks and start cloning them again. This will take about 4.5 hours, but in 2.5 hours i will know if i will have the same feature.

I will let you know what will happen.. please keep your fingers crossed..

Ceesjan

[Sajid_1]

hello,

This is the document which has been pointed by JRF:
http://support2.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000054370915

Doc ID - KBRC00000827

[Ceesjan van Hattum]

...
i reproduced the birth of this error on a clean system :(
Have a disk without LVM info (no vgcfgrestore), but vg00.conf still thinks it should have the disk.
Then rebuild /etc/lvmtab (by vgscan -v): now lvmtab okay.
And, instead of vgreduce -f, just execute a 'lvlnboot -v' to look for statusinformation...
First i should have done vgreduce -f and lvlnboot -R, but i was inpatient and did lvlnboot -v. The output of this command introduced my unknown disk "".
Now vg00 has 4 Cur PV and 3 Active PV. This 4th is the "" disk.
As i read in KBRC0000827, it is time to rebuild by hand:
vgchange -a n /dev/vg00
vgexport -m /mapfile /dev/vg00
mkdir /dev/vg00
mknod
vgimport.
But to do this, i need to boot in in Maintanance mode. But this machine is 24/7 production.. :(((

Anybody some suggestions?

[James R. Ferguson]

Hi Ceesjan:

Thanks for taking the time to post what you've reproduced.

As for the fact that this is "production"...well, schedule a downtime! :-))

Regards!

...JRF...

[Ceesjan van Hattum]

Guess what: problem solved!
What i did not tell is that this OS was 10.20. And even worse: no proper critical patches were installed.
Now i installed critical LVM and mmap patches and reconstruction of proper LVM-confs i no problem anymore.
I used: PHCO_23437 (containing critical lvlnboot patches) and PHKL_24517 (critical mmap).
And indeed: downtime needed.
Cause of problem was lvlnboot corrupting LVM structure, lvreduce causing bad blocks resuling in mmap panic.

Thanks to all of you who helped me guiding me arround.

Greetings,
Ceesjan