SUMMARY: Help! All PVs present but vgscan fails to assemble volume group
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
Everything is working once again, and I wanted to summarize my experience to the list in the hopes that someone else will find it helpful.
There appear to have been two problems at work here:
(1) The list compacting code in pv_read_all_pv_of_vg() is buggy, and in common situations would lose track of some PVs. I submitted a very simple patch for this to Heinz earlier today, but I've since put a cleaner version up at <http://larsshack.freeservers.com/lvm/>.
(2) It looks as if the 'vgmerge' command may have trodded upon on-disk information on /dev/md0, erasing both the LV table and the mappings for the first 173 PEs.
Andreas provided the information I needed to manually reconstruct the data lost due to (2). In fact, Andreas has been extraordinarily helpful in tracking down both the cause and a solution to this problem, and I'd like to thank him here for his time and effort.
I have not yet investigated the vgmerge problem in any detail, so I don't know why it corrupted the on-disk data.
-- Lars