Rabbitmq fails with error that user nova already exist on compute reinstallation

Detailed description:

During swarm ran at CI[0] with scenario
        Scenario:
            1. Revert the snapshot
            2. Create an OS volume and OS instance
            3. Mark 'cinder' partition to be preserved
            4. Mark 'vm' partition to be preserved
            5. Reinstall the compute node
            6. Run network verification
            7. Run OSTF
            8. Verify that the volume is present and has 'available' status
               after the node reinstallation
            9. Verify that the VM is available and pingable
               after the node reinstallation

there were failure [1]:
2016-04-12 23:56:40 +0000 Puppet (err): Execution of '/usr/sbin/rabbitmqctl add_user nova h69rVcAgTJheTAWzPXaRzxFF' returned 70: Creating user "nova" ...
Error: user_already_exists: nova

Also, there were an error before that[2]:
2016-04-12 23:47:37 +0000 Puppet (notice): Compiled catalog for node-2.test.domain.local in environment production in 2.73 seconds
2016-04-12 23:51:15 +0000 /Stage[main]/Rabbitmq::Management/Rabbitmq_user[guest] (err): Could not evaluate: Command is still failing after 180 seconds expired!

[0] https://product-ci.infra.mirantis.net/job/9.0.system_test.ubuntu.partition_preservation/76/console
[1] http://paste.openstack.org/show/493931/
[2] http://paste.openstack.org/show/493933/