Bug 1403355

I'm setting the same thing atomic-openshift-3.3.1.7-1

I had the customer run `rpm -qa | grep -ie openshift -ie ocp -ie ose -ie ansible` and this is the output:

openshift-ansible-roles-3.3.54-1.git.0.61a1dee.el7.noarch
openshift-ansible-filter-plugins-3.3.54-1.git.0.61a1dee.el7.noarch
atomic-openshift-master-3.3.1.5-1.git.0.62700af.el7.x86_64
atomic-openshift-node-3.3.1.5-1.git.0.62700af.el7.x86_64
ansible-2.2.0.0-1.el7.noarch
openshift-ansible-lookup-plugins-3.3.54-1.git.0.61a1dee.el7.noarch
atomic-openshift-utils-3.3.54-1.git.0.61a1dee.el7.noarch
openshift-ansible-docs-3.3.54-1.git.0.61a1dee.el7.noarch
openshift-ansible-callback-plugins-3.3.54-1.git.0.61a1dee.el7.noarch
atomic-openshift-clients-3.3.1.5-1.git.0.62700af.el7.x86_64
procps-ng-3.3.10-5.el7_2.x86_64
atomic-openshift-sdn-ovs-3.3.1.5-1.git.0.62700af.el7.x86_64
openshift-ansible-3.3.54-1.git.0.61a1dee.el7.noarch
atomic-openshift-3.3.1.5-1.git.0.62700af.el7.x86_64
tuned-profiles-atomic-openshift-node-3.3.1.5-1.git.0.62700af.el7.x86_64
openshift-ansible-playbooks-3.3.54-1.git.0.61a1dee.el7.noarch

Is the expected functionality to replace both openshift/ and rhscl/ repositories with that of oreg_url ?

@Scott So I went looking into this deeper and started digging into the ansible playbooks.

I found the default setting for the registry_url and registry_host variables [0] which are the variables used to see if ansible should replace the registry.access.redhat.com (as seen in [1]).

The example inventory file includes a block about using the oreg_url to set change the registry url [2] and I did find where it changes the registry_url [3] but I couldn't find if this is then used later to set the registry_host variable after the default. 

Could this be part of the problem?

[0] https://github.com/openshift/openshift-ansible/blob/6796a7b5bcfe391bf4affaf17c8b9a05a82be58f/roles/openshift_hosted_templates/defaults/main.yml#L7-L8

[1] https://github.com/openshift/openshift-ansible/blob/0eaa3699e7754d38b9f2b8b06d746b1cb61bc6ce/roles/openshift_hosted_templates/tasks/main.yml#L32-L35

[2] https://github.com/openshift/openshift-ansible/blob/master/inventory/byo/hosts.ose.example#L92-L96

[3] https://github.com/openshift/openshift-ansible/blob/09b3183997a65c2835a95cf8c38c28e3f20380a7/roles/openshift_node/tasks/main.yml#L25

I can reproduce it with version openshift-ansible-3.3.57-1.git.0.f275588.el7.noarch. And it has been fixed in 3.4 because of BZ#1384772.

@sdodson Does the fix in this PR (https://github.com/openshift/openshift-ansible/pull/2573) need to be back ported to 3.3?  Given that the issue was fixed in 3.4 with this PR, and the target release for this issue is 3.5 I was unsure if we needed to backport or just close this as fixed.

If we can verify that it's not a problem in 3.5 currently lets reset target version to 3.3.1 and fix it there. I'm not certain if that PR alone is sufficient to fix the problem or not.

Changes for this fix have been confirmed backported to release 3.3.

Verified with version openshift-ansible-3.3.61-1, setting oreg_url and openshift_examples_modify_imagestreams in the inventory file, installation succeed.

[root@ansible ~]# cat hosts
...
oreg_url=registry.ops.openshift.com/openshift3/ose-${component}:${version}
openshift_examples_modify_imagestreams=true
...

[root@openshift ~]# oc get pods | grep "registry-console"
registry-console-1-erbs0   1/1       Running   0          1h

[root@openshift ~]# oc get dc registry-console -o yaml | grep "image:"
        image: registry.ops.openshift.com/openshift3/registry-console:3.3