You're most probably missing some middleware sections from your registry config.yml file. Please update them according to our documentation:

https://docs.openshift.org/latest/install_config/install/docker_registry.html#docker-registry-deploying-updated-configuration

Please extend your config file at least for:

middleware:
  registry:
    - name: openshift
  storage:
    - name: openshift

Does it help?

Yes, when extend config file, the issue has fixed,thanks.

middleware:
  registry:
    - name: openshift
  storage:   
    - name: openshift
  repository:
    - name: openshift
      options:
        pullthrough: true

openshift version
openshift v3.3.0.12
kubernetes v1.3.0+57fb9ac
etcd 2.3.0+git

When we use native s3 registry config by openshift-ansible, the docker-registry's middleware stanza should contain the following by default:
  registry:
    - name: openshift
  storage:   
    - name: openshift

Will close this, and open a new.

Proposed PR: https://github.com/openshift/openshift-ansible/pull/2280

Michal,

Is that additional bit of config valid for 2.0 registry too? As in, should we add this for OSE 3.2 installs as well?

(In reply to Scott Dodson from comment #6)
> Is that additional bit of config valid for 2.0 registry too? As in, should
> we add this for OSE 3.2 installs as well?

No, it's not valid. I will be valid if we backport 2.4 registry though [1].

[1] https://github.com/openshift/ose/pull/314

So is there a document bug here?

https://docs.openshift.com/enterprise/3.2/install_config/install/docker_registry.html#docker-registry-configuration-reference-middleware

@ghunag there was :-). It's fixed now. See [1].

[1] https://github.com/openshift/openshift-docs/pull/2630

Confirmed on today's build env, the issue has fixed. 
openshift version
openshift v3.3.0.18
kubernetes v1.3.0+507d3a7
etcd 2.3.0+git


middleware:
  registry:
    - name: openshift
  repository:
    - name: openshift
      options:
        acceptschema2: false
        pullthrough: true
        enforcequota: false
        projectcachettl: 1m
        blobrepositorycachettl: 10m
  storage:
    - name: openshift

> middleware:
>   registry:
>     - name: openshift
>   repository:
>     - name: openshift
>       options:
>         acceptschema2: false
>         pullthrough: true
>         enforcequota: false
>         projectcachettl: 1m
>         blobrepositorycachettl: 10m
>   storage:
>     - name: openshift

This is a swift config file generated not by openshift-ansible. Currently openshift-ansible doesn't support swift backend storage for docker-registry, so we should check it on AWS. 

https://github.com/openshift/openshift-ansible/pull/2280 isn't in latest openshift-ansible rpm package. @Michal, we regard this bug as a installer bug. So I move to assigned first, and will check again once fixed in puddle.

Gan,

That PR should be in both 3.2.22-1 and the 3.3 version today. Marking ON_QA.

Docker registry failed to deploy. "repository" is not in the right place as we expect. I'm not sure what's wrong.(it should be in the same level as "registry" and "storage")

[root@ip-172-18-6-22 ~]# oc logs docker-registry-6-3j350
time="2016-08-15T06:42:43-04:00" level=fatal msg="Error parsing configuration file: yaml: line 28: did not find expected '-' indicator" 

Get registry-config:

<--snip-->
auth:
  openshift:
    realm: openshift
middleware:
  registry:
  - name: openshift
repository:
  - name: openshift
    options:
      pullthrough: True
  storage:
  - name: openshift

Gan,

repository section needs to be at the same level as registry and storage like this:

auth:
  openshift:
    realm: openshift
middleware:
  registry:
   - name: openshift
  repository:
   - name: openshift
     options:
       pullthrough: True
  storage:
   - name: openshift

Oh I see, that's an ansible bug. Sorry for my useless comment. Moving back to ASSIGNED.

I'm confused that I got a different config.yml today in ansible-2.2.0-0.5.prerelease.el7.noarch. Not sure the issue is related to ansible version.

<--snip-->
auth:
  openshift:
    realm: openshift
middleware:
repository:
  - name: openshift
    options:
      pullthrough: True

https://github.com/openshift/openshift-ansible/pull/2314 fixes the padding. Part of the difficulty in testing this is that the secret is not overwritten on subsequent runs. We should decide if we should do that or not in the future.

I tried it out using master branch. repository is at the right level. But looks like "openshift.common.version_gte_3_3_or_1_3" was set to false. Get the config.yml:

<--snip-->
auth:
  openshift:
    realm: openshift
middleware:
  repository:
  - name: openshift
    options:
      pullthrough: True

Remove the dependence of BZ#1368034 due to it's a edge case to set "openshift_image_tag=v3.3".

Confirmed on aws, the issue has fixed:

middleware:
  registry:
  - name: openshift
  repository:
  - name: openshift
    options:
      pullthrough: True
  storage:
  - name: openshift

openshift version
openshift v3.3.0.24-dirty
kubernetes v1.3.0+507d3a7
etcd 2.3.0+git

will verify it.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2016:1933