[Linux-cluster] LVM over a gnbd exported device

[Ion Alberdi]

Hello.
I'm currently seting up a three node cluster.
I want to have a shared logical partition between the three nodes i.e 
that the device
/dev/vg1/lv1 appears on the three nodes.

There is what I tryed
after initialing the cluster:

(ccsd 
 cman_tool join                   
 fence_tool join                  
 clvmd                            
 vgchange -aly 
 worked well)                

export /dev/hdb from the node debian to buba and gump.
debian has /dev/hdb
buba has /dev/gnbd/dd
gump has /dev/gnbd/dd

 From gump, I tried to create a logical volume from the shared device, 
everything goes well until lvcreate:

[root at gump gnbd]#pvcreate /dev/gnbd/dd
  Physical volume "/dev/gnbd/dd" successfully created
[root at gump gnbd]# vgcreate vg1 /dev/gnbd/dd
  Volume group "vg1" successfully created
[root at gump gnbd]# lvcreate -l 1500 -n lv1 vg1
  Error locking on node gump: Internal lvm error, check syslog
  Error locking on node buba: Internal lvm error, check syslog

The syslog from buba end gump were the same:

Feb 18 17:27:10 gump lvm[6634]: Volume group for uuid not found: 
6fz5zUSDvMbKnau8HdhjawXiduq1wIYoRRj2kR4Mf8rM9YE1bisYu671jD26sijR

For the debian node everything went well, because it had  /dev/vg1/lv1 
and because lvscan gave:

17:27 root at debian ~# lvscan
  ACTIVE            '/dev/vg1/lv1' [5,86 GB] inherit


On the two other nodes I had
[root at buba dev]# lvscan
  inactive          '/dev/vg1/lv1' [5,86 GB] inherit.

It seems that all worked well, except  the activation of the logical 
volume for buba and gump because they didn't find a volume groupe for 
the uuid of the shared physical device.......

Has someone any idea about fixing it?

Thanks in advance.

There is my cluster.conf:

<?xml version="1.0"?>
<cluster name="cluster1" config_version="1">
 
  <clusternodes>
    <clusternode name="buba" votes="1">
      <fence>
        <method name="single">
          <device name="human" ipaddr="200.0.0.10"/>
        </method>
      </fence>
    </clusternode>
   
    <clusternode name="gump" votes="1">
      <fence>
        <method name="single">
          <device name="human" ipaddr="200.0.0.97"/>
        </method>
      </fence>
    </clusternode>
 
    <clusternode name="debian" votes="1">
      <fence>
        <method name="single">
          <device name="human" ipaddr="200.0.0.102"/>
        </method>
      </fence>
    </clusternode>
  </clusternodes>

  <fence_devices>
    <device name="human" agent="fence_manual"/>
  </fence_devices>

</cluster>