tcp 5672 not listening for ipv4 (amqp connections fail)

[Ryan Beisner]

In a 3-node Xenial cluster, beam is not binding to tcp5672. The juju status is incorrectly OK, because the charm only checks for a pid, which does exist (bound to tcp6).

Subsequently, ipv4 connections to the rabbitmq server fail.

This impacts the rabbitmq-server next charm @master

## Amulet output
DEBUG:runner:2016-09-11 02:27:10,946 connect_amqp_by_unit DEBUG: Connecting to amqp on 172.17.104.152:5672 (rabbitmq-server/0) as testuser1...
DEBUG:runner:amqp connection failed to 172.17.104.152:5672 as testuser1 ('BlockingConnection' object has no attribute 'server_properties')

## Netstat (beam not listening on 0.0.0.0:5672)
tcp 0 0 0.0.0.0:25672 0.0.0.0:* LISTEN 15667/beam
tcp 0 0 0.0.0.0:111 0.0.0.0:* LISTEN 22863/rpcbind
tcp 0 0 0.0.0.0:4369 0.0.0.0:* LISTEN 15623/epmd
tcp 0 0 0.0.0.0:22 0.0.0.0:* LISTEN 2292/sshd
tcp 0 0 0.0.0.0:5666 0.0.0.0:* LISTEN 28753/nrpe
tcp6 0 0 :::5672 :::* LISTEN 15667/beam
tcp6 0 0 :::111 :::* LISTEN 22863/rpcbind
tcp6 0 0 :::4369 :::* LISTEN 15623/epmd
tcp6 0 0 :::22 :::* LISTEN 2292/sshd
tcp6 0 0 :::5666 :::* LISTEN 28753/nrpe

## /etc/rabbitmq/rabbitmq.config
[
    {rabbit, [

     {tcp_listeners, [5672]},

        {cluster_partition_handling, ignore}

    ]}
].

## juju status for the rmq units is ok
rabbitmq-server/0 active executing 1.25.6 4 5672/tcp 172.17.104.152 Unit is ready and clustered
  nrpe/0 unknown idle 1.25.6 172.17.104.152
rabbitmq-server/1 active idle 1.25.6 5 5672/tcp 172.17.104.154 Unit is ready and clustered
  nrpe/2 unknown idle 1.25.6 172.17.104.154
rabbitmq-server/2 active idle 1.25.6 6 5672/tcp 172.17.104.153 Unit is ready and clustered

[Ryan Beisner (1chb1n)]

Nevermind. Netstat showing tcp6 :::5672 means that tcp (4) is also listening on that port.