Red Hat Bugzilla – Bug 1255324
[RFE] connection to openstack amqp provider need to be granular
Last modified: 2016-02-19 16:39:48 EST
Description of problem:
When connecting CFME to RHEOSP7 you need to add amqp provider details. The credentials are then used against the IP in the main provider section. The problem is, with rhelosp7, the queue isn't exposed on that IP anymore. Often that IP is an external one, but in rhelosp7 queues use an internal ha mechanism and are not exposed via haproxy or a vip. Because of this you have to add the amqp port to haproxy.
Version-Release number of selected component (if applicable):
Connecting CFME to rhelosp7 overcloud.
Steps to Reproduce:
1. Deploy director based rhelosp7
2. Create overcloud (cloud) provider connection.
3. Fill our amqp details
amqp port not listening on IP added
amqp is reachable
This is due to a change in rhelosp7 director deployments to not use pacemaker or haproxy for rabbit. It would be great if cfme could be more granular in it's ability to allow you to set a separate ip for the queue (and also a unique port if required -ie if ssl, etc).
I think we need to address 2 issues regarding the rabbit setup in an OSP7 HA environment and reaching it from CFME:
1. All the 3 OSP controllers run a rabbit instance with mirrored queues. In order for CFME to use all of them in a HA manner the rabbit client needs to know about all the 3 rabbit ip addresses.
2. Rabbit binds on the internal_api network which is usually an isolated network so CFME won't be able to reach it unless it has a direct nic connected to that network.
*** This bug has been marked as a duplicate of bug 1296146 ***