[Pacemaker] All dependend resources are restarted when Clone Set changes status on ONE machine

[Sebastian Urban]

Hi.

I have two servers (cipvm1, cipvm2) and want to configure automatic
failover of virtual machines (KVM via libvirt). Both servers share a
virtual block device via DRBD in dual primary mode. I use an OCFS2
filesystem for storage of the virtual machines on top of DRBD. Pacemaker
is in full control of DRBD, OCFS2 and the virtual machines. Everything
seems to run fine.

However, let's assume that the VM "ciptest" is started on cipvm2. If
cipvm1 (not 2) fails or is put into standby then "ciptest" is stopped
and restarted on cipvm2 without any reason. I suspect that Pacemaker
thinks that it has to restart the Clone Set vmfs on cipvm2 and therefore
also restarts the VM. How can I stop Pacemaker from doing that?

Here is my config:

node cipvm1 \
        attributes standby="off"
node cipvm2 \
        attributes standby="off"
primitive ciptest ocf:heartbeat:VirtualDomain \
        params domain="ciptest" \
        meta target-role="Started" is-managed="true" \
        op stop interval="0" timeout="5m" \
        op start interval="0" timeout="5m" \
        op monitor interval="15s" timeout="5m"
primitive pri_drbd_vms ocf:linbit:drbd \
        params drbd_resource="vm_drbd" \
        operations $id="vm_drbd-operations" \
        op monitor interval="20" role="Master" \
        op monitor interval="30" role="Slave"
primitive pri_vmfs ocf:heartbeat:Filesystem \
        params device="/dev/drbd1" directory="/srv/vms" fstype="ocfs2" \
        op monitor interval="30s"
ms drbd_vms pri_drbd_vms \
        meta resource-stickiness="100" notify="true" master-max="2" \
        interleave="true" target-role="Started" is-managed="true"
clone vmfs pri_vmfs \
        meta interleave="true" ordered="true" target-role="Started"
colocation col_ciptest inf: ciptest vmfs:Started
order ord_ciptest inf: vmfs:start ciptest:start
property $id="cib-bootstrap-options" \
        dc-version="1.0.6-cebe2b6ff49b36b29a3bd7ada1c4701c7470febe" \
        cluster-infrastructure="openais" \
        expected-quorum-votes="2" \
        stonith-enabled="false" \
        no-quorum-policy="ignore" \
        last-lrm-refresh="1266089808"

Many thanks,
Sebastian