VG version 2, LVMREC02, duplicate PVID in a VG?

[lffisher]

VG 2, LVMREC02, duplicate PVIDs?

For the PVID, use:

    xd -An -j8208 -N17 -tc $MYDISK

We are using the above method to identify unique disks in a VG by using their 17 char PVID as in that post.
On a system, we have a situation with duplcate PVIDs.

We had a VG with 18 LVs.  We did a pvcreate, added the new disk to the VG with vgextend and created an LV on it.
However, what we see is that the new disk had a 17 char PVID that is identical to a disk that was already in the VG.
This does not seem to be valid.

Is this a valid situation?
Is the PVID really just those 17 chars, and should they be unique to a VG?
After those 17 chars, there is a date, is that possibly part of the VG?

How does pvcreate come up with what is supposedly a new PVID?
Is there a status file somewhere that has the last PVID used and it just adds 1?

This IS a CLUSTERED system.  We do not know if the active node has been changed or not, is it possible that the VG was created on one node, then the active switched, and the next pvcreate done on the newly active node created a bad duplicate since some status/file in the OS was different?

If this IS an invalid situation, there does not seem to be a PV change id command, similar to vgchgid.
How can this be (easily?) fixed?

Thank you, 
Larry

[Patrick Wallek]

Are you using legacy device files (/dev/dsk or /dev/rdsk) or agile device files (/dev/disk or /dev/rdisk) when looking at the PVID?

If you're using legacy device files then you  may just be seeing an alternate path to the same disk.

[lffisher]

Patrick,

Thank you for the response.
We are using agile, /dev/disk format.
We are sure it is not the same disk, it is a separate disk that was added to the VG, and got pvcreated with a PVID the same as a disk already in the VG.

For VG version 2, how does pvcreate come up with a unique PVID?  Because it seems it didn't, assuming the PVID is just those 17 chars.  Maybe it is related to the machine being in a cluster?  I don't have the history, but maybe the first PV created in the past on another machine in the cluster, and then a year later running pvcreate on a different cluster machine duplicated a PVID?  

If there is a dup PVID in this VG, can we just pvcreate another new disk, pvmove from the "bad" disk (with the dup PVID) to this new disk, and then remove the "bad" disk?

Seems that HPUX shouldn't have allowed the second disk with the same 17 char PVID to be added to the VG, assuming we are interpreting what happened properly.

Thank you.

[liuyl_it]

Exactly,  there should be no any duplicated PVIDs allowed in the same VG during its any alteration with either v1.0 or v2.0, except for its pvlinks！