Hide Forgot
Description of problem: The test registry was deleted during upgrade, i add an registry in docker config for ose3.3 image, but it will be set to default registry only(registry.access.redhat.com) during the upgrading, then result that upgrade fail because no v3.3.x image could be found in default registry.access.redhat.com. Version-Release number of selected component (if applicable): atomic-openshift-utils-3.3.13-1.git.0.7435ce7.el7.noarch How reproducible: always Steps to Reproduce: 1.install ose-3.2 in AH726. 2.update installer to 3.3. 3.add an registry to /etc/sysconfig/docker ADD_REGISTRY='--add-registry qa.test.registry.redhat.com --add-registry registry.access.redhat.com' INSECURE_REGISTRY='--insecure-registry qa.test.registry.redhat.com' 4.specify openshift_image_tag in inventory file. ... [OSEv3:vars] ansible_ssh_user=cloud-user ansible_become=yes deployment_type=openshift-enterprise openshift_image_tag=v3.3.0.22 ... 5.ansible-playbook -i hosts /usr/share/ansible/openshift-ansible/playbooks/byo/openshift-cluster/upgrades/v3_3/upgrade.yml Actual results: it upgraded fail,i checked the docker config file and found that the add_reigstry which i configured in step 3 has been overwrited. ADD_REGISTRY='--add-registry registry.access.redhat.com' Expected results: The existing docker config should be merged with the defined parameters in inventory hosts file. It should not overwrite the registry directly which be configured by users. Additional info:
If using atomic-openshift-installer I think you will have to set the OO_INSTALL_ADDITIONAL_REGISTRIES and OO_INSTALL_INSECURE_REGISTRIES environment variables. If using ansible directly you can set the following in your inventory: openshift_docker_additional_registries openshift_docker_insecure_registries
Agreed, upgrade uses the docker role which has to generate the registry lines in /etc/sysconfig/docker. If you're using an inventory and ansible directly (like your steps above) you can configure these with: openshift_docker_additional_registries=brew-pulp-docker01.web.prod.ext.phx2.redhat.com:8888 openshift_docker_insecure_registries=brew-pulp-docker01.web.prod.ext.phx2.redhat.com:8888 If you're using the quick installer, Brenton's env variables method should still work and is likely the best option. You could also set the above vars in the deployment of a quick installer config file but that would require generating a quick installer config, then modifying it, and re-running, so the env var method Brenton suggests is probably best.
Many thx. I've got your meaning. It is indeed not a bug but a misunderstanding about the difference between quick installer and ansible install.