[linux-lvm] Strange lvm on raid1 on top of multipath problem

[Rainer Krienke]

Hello,

I'am in trouble running lvm (lvm-1.0.6 on a suse8.2 system with kernel 2.4.20) 
on a raid md device which in turn is based on two md multiptah devices:
 
                         /dev/md20  (raid1)
              /dev/md10 (mp)     /dev/md13 (mp)   mp=multipath
    	         disk1                              disk2

cat /proc/mdstat says this (just to make things clearer):

md20 : active raid1 md10[0] md13[1]
      903371648 blocks [2/2] [UU]

md13 : active multipath sde2[0] sdh2[1]
      903371712 blocks [2/2] [UU]

md10 : active multipath sdd1[0] sdg1[1]
      903373696 blocks [2/2] [UU]

The basic setup worked just fine. I created one physikal volume on /dev/md20 
(800GB) then one volumegroup "data" and then several logical volumes. So far 
everything was fine. Then I deleted one logical volume and trouble started. 
After the deletion I can no longer run vgscan. It keeps telling me (please 
see attachment for vgscan -d output):

vgscan -- reading all physical volumes (this may take a while...)
vgscan -- found active volume group "data"
vgscan -- ERROR "pv_check_consistency_all_pv(): PE" volume group "data" is 
inconsistent
vgscan -- ERROR: unable to do a backup of volume group "data"
vgscan -- ERROR "lvm_tab_vg_remove(): unlink" removing volume group "data" 
from "/etc/lvmtab"
vgscan -- "/etc/lvmtab" and "/etc/lvmtab.d" successfully created
vgscan -- WARNING: This program does not do a VGDA backup of your volume group

I noticed that something must have gone wrong with the physical volume. It 
seems to me that lvm did not only recognice the physical volume on /dev/md20 
but somehow also on the underlying sub mirror devices /dev/md10 and 
/dev/md13. Right after I created the logical volumes lvmdiskscan showed this:
...
lvmdiskscan -- /dev/md10   [     861.52 GB] free meta device
lvmdiskscan -- /dev/md13   [     861.52 GB] free meta device
lvmdiskscan -- /dev/md20   [     861.52 GB] USED LVM meta device
...

Since the deletion of the logical volume it says:

...
lvmdiskscan -- /dev/md10   [     861.52 GB] USED LVM meta device
lvmdiskscan -- /dev/md13   [     861.52 GB] USED LVM meta device
lvmdiskscan -- /dev/md20   [     861.52 GB] USED LVM meta device
...

Is there any known problem with lvm on raid1 on top of multipath devices? 
Could it be that lvm wrote a pv-signature not only one the real physical 
volume on /dev/md20 but also on /dev/md10 and /dev/md13 or is this a suse 
bug?

I have to note that I did the changes to the PVs, VGs, LVs all with yast from 
suse not with the pv* vg* lv* commandline tools. Is this a know source of 
trouble?

Would  be very greatful for any help since the system in question actually 
should go into production very soon.....

Thanks Rainer
-- 
---------------------------------------------------------------------------
Rainer Krienke, Universitaet Koblenz, Rechenzentrum
Universitaetsstrasse 1, 56070 Koblenz, Tel: +49 261287 -1312, Fax: -1001312
Mail: krienke at uni-koblenz.de, Web: http://www.uni-koblenz.de/~krienke
Get my public PGP key: http://www.uni-koblenz.de/~krienke/mypgp.html
---------------------------------------------------------------------------
-------------- next part --------------
A non-text attachment was scrubbed...
Name: vg.gz
Type: application/x-gzip
Size: 5121 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/linux-lvm/attachments/20030530/12dcf50f/attachment.bin>