Description of problem:

Even though restart is specified to recovery policy in a service, the servie is relocated.
About cluster.conf see the attachment.


Version-Release number of selected component (if applicable):

rgmanager-2.0.46-1.3.

How reproducible:

The service is httpd. Stopping the httpd with service command.

Steps to Reproduce:
1. Deploy cluster with attach cluster.conf
2. Do service httpd stop at host where the httpd run.
3.
  
Actual results:

The service migrate to another cluster node without any trying to restart.

Expected results:

Restart httpd again at the node where it has run.

Additional info:

After applying the attached patch, the restart policy works as expected.