rabbitmq hostname resolution error while deploying

[Victor Tapia]

rabbitmq-server does not resolve its machine's hostname and is unable to start the service.

How to reproduce:

- Deploy by running:

> juju-deployer --bootstrap -c default-ceilometer-ha-only.yaml -d trusty-icehouse
(rev10 yaml from lp:~openstack-charm-testers/+junk/ceilometer-ha/ )

Workaround:

- Add an entry to /etc/hosts, such as:
127.0.1.1 juju-vtapia-machine-20
- Start the service and resume the deployment

[Edward Hope-Morley (hopem)]

Victor could you provide a bit more info about the environment that you are deploying in? This should only be a problem if (a) you are not using ipv6 mode in that charms and (b) your rabbitmq-server node is unable to resolve name like juju-vtapia-machine-20 from the configured dns server. This *should* work fine if you have dns setup properly but I think there are some cases (MaaS dns?) where the name does not get resolved properly.

[Edward Hope-Morley (hopem)]

So it seems that this should no longer be a problem with the latest version of MaaS 1.7. Could you please re-test?

[Victor Tapia (vtapia)]

Just to complete the original info with some logs:

- The deployment fails with:

2014-12-12 11:43:01 [DEBUG] deployer.env: Delta unit: rabbitmq-server/0 change:error
2014-12-12 11:43:01 [ERROR] deployer.env: The following units had errors:
   unit: rabbitmq-server/0: machine: 20 agent-state: error details: hook failed: "config-changed"
2014-12-12 11:43:01 [INFO] deployer.cli: Deployment stopped. run time: 1262.45

- The rabbitmq-server node shows the following:

ubuntu@juju-vtapia-machine-20:~$ sudo service rabbitmq-server status
Status of node 'rabbit@juju-vtapia-machine-20' ...
Error: unable to connect to node 'rabbit@juju-vtapia-machine-20': nodedown

DIAGNOSTICS
===========

nodes in question: ['rabbit@juju-vtapia-machine-20']

hosts, their running nodes and ports:
- juju-vtapia-machine-20: [{rabbitmqctl22188,47551}]

current node details:
- node name: 'rabbitmqctl22188@juju-vtapia-machine-20'
- home dir: /var/lib/rabbitmq
- cookie hash: MnDiPHIglyN5Lj42a4v1+w==

ubuntu@juju-vtapia-machine-20:~$ sudo service rabbitmq-server start
 * Starting message broker rabbitmq-server * FAILED - check /var/log/rabbitmq/startup_\{log, _err\}
                                                                                                                                                                                                     [fail]
ubuntu@juju-vtapia-machine-20:~$ cat /var/log/rabbitmq/startup_log
ERROR: epmd error for host "juju-vtapia-machine-20": address (cannot connect to host/port)

-

[Victor Tapia (vtapia)]

Currently we're using MAAS 1.6. Once upgraded, I'll re-test

[James Page (james-page)]

Does juju-vtapia-machine-20 resolve to a single IP address?

[James Page (james-page)]

I think this has been resolved as part of the recent refactoring of the rmq charm to avoid having to know the hostname of the host via DNS.

Marking as 'Fix Released'; please re-open if this is not the case.