[Linux-cluster] unable to start virtual service

[van Schelve]

Hello.

I have just set up a test cluster environment based on RHEL5.1 advanced
platform + latest updates with two nodes + shared storage. What I want to
test is cluster services for virtual maschines because we need live
migration with xen.

The cluster itself is OK. I also have access to shared storage from both
nodes.

On each node I have a mountpoint /var/xen where the configurations for the
virtual maschines are located:

/dev/mapper/san01VG-xenconfiglv on /var/xen type gfs
(rw,hostdata=jid=0:id=196610:first=0)

As disk for the VM I have a lv based on clvm:

disk = [ "phy:/dev/san01VG/vux002-01-lv,xvda,w" ]

I have configured the virtual service with lucie and I am able to start the
virtual maschine manually using xm create vux002. I also can do a live
migration from node1 to node2 and back manually or with luci. But I cannot
start or enable the virtual maschine with luci. When I try to do this I get
every time a generic error:

Feb 23 11:31:11 ux071 clurgmgrd[28779]: <notice> Recovering failed service
vm:vux002 
Feb 23 11:31:11 ux071 clurgmgrd[28779]: <notice> start on vm "vux002"
returned 1 (generic error) 
Feb 23 11:31:11 ux071 clurgmgrd[28779]: <warning> #68: Failed to start
vm:vux002; return value: 1 
Feb 23 11:31:11 ux071 clurgmgrd[28779]: <notice> Stopping service vm:vux002

Feb 23 11:31:17 ux071 clurgmgrd[28779]: <notice> Service vm:vux002 is
recovering 
Feb 23 11:31:22 ux071 openais[28485]: [TOTEM] Retransmit List: 1be  

Is it possible that virtual service can only start a virtual maschine when
the disk is in a image file (by the way this would be the same behavior as
with the satellite)?

If it's so it would be very bad. Using image files for the virtual
maschines have extreme impact on performance and stability of the virtual
maschine. 

Is there a workaround of a fix available?

Thank you

-- Hans-Gerd van Schelve