kolla-ansible pull with kolla_ansible-4.0.0.0rc1 fails, because of missing tag in docker registry

[David Rabel]

I just tried to do an all-in-one deployment on a bare centos7. It fails when pulling with the following error:

TASK [memcached : Pulling memcached image] *************************************
fatal: [localhost]: FAILED! => {"changed": false, "failed": true, "msg": "Unknown error message: Tag 4.0.0 not found in repository docker.io/kolla/centos-binary-memcached"}

The tag is indeed missing in dockerhub.

I installed kolla-ansible via pip, so it's version 4.0.0rc1

[Eduardo Gonzalez (egonzalez90)]

Docker images with tag 4.0.0 are not pushed to Dockerhub yet.
https://hub.docker.com/r/kolla/centos-source-memcached/tags/
Final Ocata release is not published for kolla and other release:cycle-trailing model projects.
Now is up for testing before release next week. Once released docker images will be pushed to dockerhub tho.
Check release calendar https://releases.openstack.org/ocata/schedule.html#o-finalrc

[David Rabel (rabel-b1)]

Hi Eduardo,

thanks for the information.

I would still consider this to be a bug, since the deployment from dockerhub doesn't work now with the included example configuration.

Also this is probably going to happen again every time a new version is released.

[David Rabel (rabel-b1)]

Jeffrey, please don't just close this bug without any comment. As I explained in my last post, I still consider this a bug.

[Steven Dake (sdake)]

Kolla's push policy is completely inconsistent. We push main releases, but not release candidates. We push master, although that may end once Ocata releases in speaking to portdirect. We push 3.0.3-beta1 release candidates for 3.0.3 to enable kolla-kubernetes. I think in general we need to have a unified view of our push policy, so keeping this bug open until this is further discussed at a team meeting to which I've added to the agenda.

[Jeffrey Zhang (jeffrey4l)]

we should not push unreleased images to hub.docker.com

unreleased images may have lots of issue and different kolla-ansible. And it is hard for end-user to find the right release of kolla-ansible.

[David Rabel (rabel-b1)]

I understand that.

Maybe for my special problem it would be sufficient to describe a workaround in the docs?
Because at the moment when you read https://docs.openstack.org/developer/kolla-ansible/quickstart.html and try to deploy an all-in-one it won't work without some extra tweaking.

[Jeffrey Zhang (jeffrey4l)]

In the doc, it says

> Otherwise, the Docker Hub Image Registry contains all images from each of Kolla’s major releases. The latest release tag is 3.0.2 for Newton.

which line do u think is bad? Then feel free to improve it.

[David Rabel (rabel-b1)]

I don't know. This is quite frustrating.

The guide is not actually wrong, there is no bad lines. It just does not explain, that users who install the most recent version of kolla-ansible via pip may have to make some changes if they want to use the docker hub image registry.

Maybe this is nit-picking. I just thought that a quick-start-guide is meant to work out of the box, so that new users can try kolla-ansible without diving too deep into it.

If you don't consider this a problem, than that's how it is. I won't mention it anymore.

[lamisse (lamiss)]

i got this one

TASK [memcached : Pulling memcached image] *************************************
fatal: [localhost]: FAILED! => {"changed": false, "failed": true, "msg": "Unknown error message: unauthorized: authentication required"}
        to retry, use: --limit @/usr/share/kolla-ansible/ansible/site.retry

PLAY RECAP *********************************************************************
localhost : ok=13 changed=0 unreachable=0 failed=1

Command failed ansible-playbook -i /usr/share/kolla-ansible/ansible/inventory/all-in-one -e @/etc/kolla/globals.yml -e @/etc/kolla/passwords.yml -e CONFIG_DIR=/etc/kolla -e action=pull /usr/share/kolla-ansible/ansible/site.yml