[Pacemaker] Colocation constraint to External Managed Resource

[Robert H.]

Hello list,

I have a question regarding colocation.

I have an external managed resource (not part of pacemaker, but running 
on the pacemaker nodes as multi master application) - in this case 
XtraDB Cluster. I also want to keep this ressource manually managed.

For accessing the XtraDb Cluster I use a failover IP "DATABASE_IP" 
(IPaddr2) (I know, I'm actually using XtraDB as a MySQl Master / Slave 
replication alike with Master / Master handling).

Now I want to have a configuration, where Pacemaker migrates the 
DATABASE_IP away from the node if the externally managed XtraDB is 
stopped. However Pacemaker should not manage the XTraDB ressource for 
start / stopp.

I tried configuring as lsb:mysql resource with start and stop 
operations "enabled=false" -> did not work.
I tried configuring as lsb:mysql resource with "is-managed=false" and 
failure-timeout="30s" -> did work, until I rebootet one of the machines. 
Then the resource was not discovered on the node after the reboot (it 
was shown as "stopped", while it was started already).

I'm using CentOS6 Pacemaker (1.1.8-7.el6-394e906) on corosync.

Any hints how to configure this ?

Regards,
Robert

--
Robert