vgcfgbackup error when extending a volumegroup

[Torbjorn Eriksson]

When trying to extend an existing volumegroup with one disk i get following error, but the disk seems to be added. The /etc/lvmconf file is not updated either.

vgcfgbackup: IO error on Physical Volume device /dev/rdsk/c4t0d1: I/O error

If I then try to reduce the volumegroup there is no error and the /etc/lvmconf file is being updated as it should.

[Kent Ostby]

Is it being updated with that disk or do you get that error any time the disk is present in the volume group ?

My guess would be that we can't read the part of the disk that contains the lvm headers.

To verify if this is a physical disk read problem, set up a window to do this:

dd if=/dev/rdsk/c4t0d1 of=/dev/null bs=64k

[RAC_1]

diskinfo /dev/rdsk/c4t0d1
Does sit return OK??
Run a dd test on the disk.

if=/dev/rdsk/c4t0d1 of=/dev/null bs=1k

Anil

[Florian Heigl (new acc) · Accepted solution]

I assume that c4t0d1 is the disk You're trying to add?

You might get that error if the VGRA of the disk is not properly initialized - maybe try first a pvcreate -f /dev/rdsk/c4t0d1 and if that doesn't work out, use a dd if=/dev/zero of=/dev/rdsk/c4t0d1 bs=1024k count=2m to completely wipe it.

[Torbjorn Eriksson]

No,c4t0d1 is the existing(active) disk in the volumegroup. Right now I'm trying to read the entire disk. It takes some time...

I have just tried to add new disk to the existing vg. If I try to do a vgcfgbackup command dirtekt on the commandline it works fine.

[Dan Yeatman]

This problem sounds like a disk hardware problem, any scsi errors in syslog for that device? What type of drive?

Which version of HPUX are you running?

What LVM cumulative patch version is installed?

There are fixes for such a problem in the LVM cummulative patches in 10.20 and 11.00.

Any chance the VGRA is not where it should be/has been relocated on that disk?

Does the problem persist after a reboot? ie could there be a timestamp issue on the VG headers.

[Torbjorn Eriksson]

I have now read both disks with dd without problems. There is no errors in the log. Both the old (active) device and the new one are SAN disks on an EMC DMX Symmetrix.

We are running HPUX 11.23

As I see it with swlist command, we don't have any LVM patches applied.

We haven't tried to reboot yet. It's a production machine.

[Torbjorn Eriksson]

I tried to wipe the hole disk with if=/dev/zero of=/dev/rdsk and then got it in to the vg again. Now it did not come with a vgcfgbackup error.

[Devender Khatana]

Hi,
It means the I/O problem is not there now. But this also means that your media may not be reliable. Just avoiding using this for critical data as it might come back again any time.

Just a thought.

HTH,
Devender