Comment 2 for bug 1480282

Well, the problem is the following:

* In 6.1 we have the /etc/yum.repos.d/auxiliary.repo points to file:///var/www/nailgun/2014.2.2-6.1/centos/auxiliary/
* When we perform a master node upgrade from 6.1 to 7.0, the puppet manifest replaces the /etc/yum.repos.d/auxiliary.repo to points it to file:///var/www/nailgun/2015.1.0-7.0/centos/auxiliary/
* In case of upgrade fail the rollback procedure is initiated.
* When we start upgrade procedure once again, it fails because yum is unable to fetch file:///var/www/nailgun/2015.1.0-7.0/centos/auxiliary/ repo (it was removed during rollback procedure).

I don't know how to fix it's better. I see two way to fix it:

1/ Move auxiliry repo management to fuel_upgrade. That means, we shouldn't create auxiliary.repo file in puppet manifests, and should create/remove this file in fuel_upgrade script.

2/ Create a 7.0_auxiliary.repo in puppet manifests, and add hack to remove this file in fuel_upgrade script (similar to this one https://github.com/stackforge/fuel-web/blob/master/fuel_upgrade_system/fuel_upgrade/fuel_upgrade/engines/host_system.py#L158-L163 )

I think anyway, in case of upgrading we should create not "auxiliary.repo", but "7.0_auxiliar.repo", just like we do for the "nailgun.repo".