Update VG on a node of a cluster

[Dany De Bontridder]

Hi,

I have 4 nodes in my cluster, someone has added a disk to a VG on node 1, so I've tried to update the definition on others nodes
vgexport -s -p -v -m vg1 /dev/vg1
and on others nodes
vgimport -v -s -m vg1 /dev/vg1
But I get the message
Beginning the import process on Volume Group "/dev/vg1".
vgimport: Volume group "/dev/vg1" already exists in the "/etc/lvmtab" file.

It is normal because the VG does exist on others nodes but I would like to have the added disk on them

example:

Node 1 : VG1 disk /dev/pv1 /dev/pv2
Node 2 : VG1 disk /dev/pv1

Thanks for you help

D.

[Steven E. Protter]

Shalom,

Some reason the volume group vg1 seems to already be in use on those systems.

It would be a good idea to find out what its being used as and pick a new name for this new volume group if its to be shared by more than one node.

This is now a naming convention annoyance that could quickly get out of control during an upgrade or expansion of cluster functionality.

SEP

[Dany De Bontridder]

Hi,

Bear with me, vg1 pv1 are fake ;-) Just to help the reading.

Actually, the VG were created with the cluster, 2 years after, some disks have been added but the vg definition is not the same everywhere.

[nishith · Accepted solution]

Hi Dany,

On the other nodes you have do "vgexport" first to remove the VG from /etc/lvmtab & then do an "vgimport".

#vgexport /dev/vg1
#mkdir /dev/vg1
#mknod /dev/vg1/group c 64 0x010000
#vgimport /dev/vg1 /dev/pv1 /dev/pv2
(careful with PV names as it need not be the same on all nodes.).

regds,
nishith.

[Dany De Bontridder]

So on node1
vgexport -s -p -m map-vg1 -s /dev/vg1
rcp map-vg1 to node2

on node 2 I do
vgexport /dev/vg1 to destroy it
mkdir && mknod
vgimport -s -m map-vg1 /dev/vg1

And... I see too many disk due to the PowerPath :-(

Any idea to fix it ?

Thanks

[Dany De Bontridder]

Ok, I've found how to I've needed to specify the PV. And because I'm using PowerPatch, I see the same disk twice,

vgexport -s -p -m vg1.map /dev/vg1
rcp node2:/tmp/vg1.map

on node2
vgexport vg1 (to destroy it)
mkdir /dev/vg1
mknod /dev/vg1/group c 64 xx0000 (replace xx by 2 digits)
syminq |grep txdx (if disk are c32t0d1, takes t0d1)
and after use only one disk for each powerpatch patch

and vgimport -s -v -m /tmp/vg1.map /dev/c32t0d1


That works :-)