Old container started during the fuel update

[Alexander Kislitsky]

Upgrading Fuel from 5.0 to 5.1.
Set of containers are not started due to rabbitmq container is not started.
In the 'supervisorctl' console we have BACKOFF error for rabbitmq container, but in 'docker ps' output we have rabbitmq container of 5.0 version. This bug can be with any container.

By the logs we can restore workflow:

1. all containers are stopped by supervisord api call
2. all containers are stopped by docker api call (due to supervisord bugs)
3. supervisord starts one of stopped containers
4. upgrading workflow
5. new container can't be started with error: Error: Cannot start container fuel-core-5.0-rabbitmq: listen tcp 0.0.0.0:15672: bind: address already in use

For fix this bug we should switch configs of containers to new version after db is dumped, cobbler configs are saved and before stopping services by supervisord api call.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/112892

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/112892
Committed: https://git.openstack.org/cgit/stackforge/fuel-web/commit/?id=fbfac1215500515ed0e7ea2648cb5b5067b3db3d
Submitter: Jenkins
Branch: master

commit fbfac1215500515ed0e7ea2648cb5b5067b3db3d
Author: Alexander Kislitsky <email address hidden>
Date: Fri Aug 8 18:23:16 2014 +0400

    Supervisord can start already stopped container

    We stop containers via supervisord api call.
    Also we try to stop containers via docker api call due to
    supervisord bugs. Sometimes supervisord starts stopped
    container again. And new container can't be run due to
    old one container already run. For workaround this supervisord
    bug we switch configs before upgrade.

    Change-Id: I595ac1cdd2c1d6fef02f6a7650d65cf968a7b850
    Closes-Bug: #1354465