vgcfgbackup

[Tammy Liang]

I got following message when I run vgcfgbackup:
vgcfgbackup: /etc/lvmtab is out of date with the running kernel:Kernel indicates 5 disks for "/dev/vg01"; /etc/lvmtab has 4 disks.
Cannot proceed with backup.

I checked /etc/lvmtab shows there are 4 disks
which is correct. How can I change kernel to
recognize there are only 4 disks instead of 5?

Thanks for any help

[Helen French]

A good explanation and solution can be found from this document TKB# KBRC00000632:

http://www1.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000065011231

[Tammy Liang]

Thank you for the doc. I checked it, but it
does not apply to my situation. I don't have
missing disk device. The number of disk are correct in /etc/lvmtab and /etc/lvmconf/vg01.conf files (4 disks). It also correct after run "vgcfgrestore -n /dev/vg01 -l" command. vgdisplay do indeed shows Cur PV is 5, and Act PV is 4. I need to fix it, so Cur PV also shows 4 instead of 5. I have too many disks instead of missing disks.

Thank you....

[Helen French]

I would try the following:

1) Check and apply all latest LVM and Kernel patches.

2) Reactivate VG (vgchange -a n and vgchange -a y)

[Marco Santerre]

I did have something similar to that that occured to me not too long ago, and it had to do with the fact that I had BCV synchronized with my standard devices and that I added a disk in my Volume Group on my standard and the split. Then my BCV Volume groups would show a discrepency between my Act PV and my Cur PV. Would you happen to have done something similar to that?

[Ashwani Kashyap]

Make a copy of the lvmtab file first .
THen run vgscan ( man vgscan ) in preview mode to see if you get the correct number of PV in the volume groups .

if you are satisfied with that , then re run vgscan without the preview option , whcih will recreate your lvmtab file . Then run vgcfgbackup again .

If it doesnot run , you can always restore your copied lvmtba file .

[Dietmar Konermann · Accepted solution]

Tammy,

actually your VG physically consists of 5 disks, believe me. The 5th (maybe phantom) disk can usually be vgreduce'ed out of the VG using:

vgreduce -f /dev/vg01

However, if LVM believes to have extents on that PV then this command would not work, but at least you get more precise information about the actual problem.

So please run the command above and post the results.

Best regards...
Dietmar.

[Rajeev Shukla]

I had the same problems few days back, what happens is that by some reasons the VGRA gets corrupted in all the disks, and if you have tools to look at those areas on the disk you'll find that it has an counter saying the number of disks involved in this VG is 5 but thats not the case if you look at you'r lvmtab.
I tried vgexport and vgimport but trust me nothing will work.
The only solution to decrement the counter on the disk is to do a
vgreduce -f /dev/vg01
after doing this you'll find Current PV and Active PV same.

[Timothy P. Jackson]

Hi Tammy,

I have seen this quite a bit in cases where a hard drive will crash and a replacement is put back into the VG incorrectly. It thinks that there are five but you really only have four. I have used Dietmar's proceedure and then followed it up by doing a vgscan to create a new lvmtab. besure to make a copy of the original lvmtab first

[Tammy Liang]

Can I run "vgreduce -f /dev/vg01" on a running production system?