[10.0 swarm] task "rabbitmq" fails

[Sergey Novikov]

Detailed bug description:

The issue was found by https://product-ci.infra.mirantis.net/job/10.0.system_test.ubuntu.custom_hostname/195/testReport/(root)/default_hostname/default_hostname/

Steps to reproduce:
            1. Create a cluster
            2. Add 3 nodes with controller role
            3. Add 1 node with compute role
            4. Deploy the cluster

2017-02-27 01:32:22 INFO [17257] Cluster[]: All nodes are finished. Failed tasks: Task[rabbitmq/5] Stopping the deployment process!

Additional info: rabbitmq cluster seems alive http://paste.openstack.org/show/600641/

And this bug doesn't look similar to https://bugs.launchpad.net/mos/+bug/1626933 - rabbitmq-server has version 3.6.6-1~u16.04+mos1

[Alexey Lebedeff (alebedev-a)]

puppet manifest was adding rabbit users right at the moment when pacemaker decided to restart some rabbits. Funny thing is that adding users in puppet is a useless operation in presence of pacemaker: those created users will be lost during resets/joins performed by OCF script.

We need to disable this user-creation activity completely - and the ONLY thing that puppet should do is to install package and drop 2 config files into their proper locations (i.e. no user management, no (re)starting/stopping/enabling of systemd unit, etc.)

[Alexey Lebedeff (alebedev-a)]

Should be fixed by https://review.openstack.org/#/c/447420/

[Alexey Lebedeff (alebedev-a)]

I've deployed an env manually with this patch and there were no traces of spurious rabbitmq restarts

[Alexey Lebedeff (alebedev-a)]

To check whether patching had an effect you need to run the following command on any controller:

crm resource param p_rabbitmq-server show host_ip

And it should return "127.0.0.1" here.