Solved my problem
First error was in constraint: i've put constraint with "cxml" resource
alone, not with cloned "cxml-clone" - that's why "cxml" were moved first
on "standby" command. after redefining constraint to "cxml-clone than
testVM" putting active node in standby went smooth - VM moved correctly,
no errors.
Second problem was because of "libvirt-guests" service, that is
suspending my VM's on host reboot command. "chkconfig libvirt-guests
off" command isnt enough, as it leaves symlinks "K01libvirt-guests" in
/etc/rc.d/rcX.d . Removing that symlinks from rc3.d and rc6.d solved
problem - now reboot process started with pacemaker shutdown and
resources moved correctly to other nodes.
Yuriy Demchenko
On 04/10/2013 02:59 PM, Yuriy Demchenko wrote:
Hi,
I've set up 3-node cluster (2 active nodes + 1 standby for quorum),
cman+pacemaker
Resources - "cxml-clone" gfs2 filesystem (cloned, run on both nodes)
and "testVM" via heartbeat:VirtualDomain (domain xml located on gfs2
fs, cLVM disk backend), set up constraints: "cxml-clone" started first
than "testVM" (symmetrical, according to description it'll be stopped
in reverse order).
While manual migration of VM runs fine (pcs resource move testVM
node-2/node-1) - succesfull live migration, VM runs uninterrupted, but
when I'm trying to reboot/put in standby node running VM - everything
is crashing, migration fails and node fenced.
From logs i can see that resource "cxml" stopped first (or
simultaneously, at least not waiting for vm migration to complete),
then migration fails because xml not available.
Apr 10 14:03:20 node-2 lrmd[2679]: notice: operation_finished:
cxml_stop_0:3282 [ 2013/04/10_14:03:20 INFO: Running stop for
/dev/cstore/cxml on /mnt ]
Apr 10 14:03:20 node-2 lrmd[2679]: notice: operation_finished:
cxml_stop_0:3282 [ 2013/04/10_14:03:20 INFO: Trying to unmount /mnt ]
Apr 10 14:03:20 node-2 lrmd[2679]: notice: operation_finished:
cxml_stop_0:3282 [ 2013/04/10_14:03:20 INFO: unmounted /mnt
successfully ]
Apr 10 14:03:20 node-2 crmd[2682]: notice: process_lrm_event: LRM
operation cxml_stop_0 (call=77, rc=0, cib-update=37, confirmed=true) ok
Apr 10 14:03:21 node-2 lrmd[2679]: notice: operation_finished:
testVM_migrate_to_0:3281 [ 2013/04/10_14:03:20 INFO: testvm: Starting
live migration to node-1 (using remote hypervisor URI
qemu+ssh://node-1/system ). ]
Apr 10 14:03:21 node-2 lrmd[2679]: notice: operation_finished:
testVM_migrate_to_0:3281 [ error: Requested operation is not valid:
domain is already active as 'testvm' ]
Apr 10 14:03:21 node-2 lrmd[2679]: notice: operation_finished:
testVM_migrate_to_0:3281 [ 2013/04/10_14:03:21 ERROR: testvm: live
migration to qemu+ssh://node-1/system failed: 1 ]
Apr 10 14:03:21 node-2 crmd[2682]: notice: process_lrm_event: LRM
operation testVM_migrate_to_0 (call=75, rc=1, cib-update=38,
confirmed=true) unknown error
Apr 10 14:03:21 node-2 lrmd[2679]: notice: operation_finished:
testVM_stop_0:3392 [ 2013/04/10_14:03:21 ERROR: Configuration file
/mnt/testvm.xml does not exist or is not readable. ]
But wtf?! I've set up constraint, and "testVM" should be stopped/moved
first, not "cxml"
What is wrong with my configuration, am I missing something?
logs and CIB in attach
_______________________________________________
Pacemaker mailing list: Pacemaker@oss.clusterlabs.org
http://oss.clusterlabs.org/mailman/listinfo/pacemaker
Project Home: http://www.clusterlabs.org
Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
Bugs: http://bugs.clusterlabs.org
