Note: This is a beta release of Red Hat Bugzilla 5.0. The data contained within is a snapshot of the live data so any changes you make will not be reflected in the production Bugzilla. Also email is disabled so feel free to test any aspect of the site that you want. File any problems you find or give feedback here.
Bug 1516534 - installer oreg_url doesn't serve as universal url - needs to
Summary: installer oreg_url doesn't serve as universal url - needs to
Alias: None
Product: OpenShift Container Platform
Classification: Red Hat
Component: Installer
Version: 3.7.0
Hardware: Unspecified
OS: Unspecified
Target Milestone: ---
: 3.10.0
Assignee: Michael Gugino
QA Contact: Johnny Liu
: 1516535 1531949 1558306 1573904 1575994 1576826 1604714 (view as bug list)
Depends On:
Blocks: 1588291 1623602
TreeView+ depends on / blocked
Reported: 2017-11-22 20:37 UTC by Erik M Jacobs
Modified: 2018-11-29 12:13 UTC (History)
30 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
: 1588291 1623602 (view as bug list)
Last Closed: 2018-08-27 20:52:26 UTC
Target Upstream Version:

Attachments (Terms of Use)

System ID Priority Status Summary Last Updated
Red Hat Knowledge Base (Solution) 3538711 None None None 2018-07-22 06:53:30 UTC

Description Erik M Jacobs 2017-11-22 20:37:28 UTC
The installer's hosts/config file takes an argument, oreg_url, that appears like it should set the image format string for all components. It does not.

For example, here's how you have to override/set all of the relevant components, in spite of setting oreg_url:${component}:${version}
# Enable service catalog
# service broker(s)
# Enable template service broker
# Enable ansible service broker

As you can see:

* You have to set "additional registry" for things that you don't give an explicit prefix
* You have to set oreg_url for anything that might happen later that the installer didn't explicitly tell openshift to do
* you have to override many different prefixes either explicitly or partially (to pick up additional registry)

setting oreg_url should automatically override all component installations without needing to additionally dig into component prefixes. It should be optional whether or not to need to override additional prefixes.

Comment 1 Scott Dodson 2018-01-22 17:42:21 UTC
*** Bug 1531949 has been marked as a duplicate of this bug. ***

Comment 2 Nicholas Nachefski 2018-02-10 18:40:31 UTC
In case anyone else needs help with this, here is a comprehensive list of variables that need to be set for disconnected installs to work.  Setting all of these is a work around until the openshift-ansible playbooks get normalized.

These work for 3.7 and 3.9.,${component}:${version}

Comment 3 Johnny Liu 2018-03-20 06:18:15 UTC
Some related github issue:

Comment 4 Johnny Liu 2018-03-20 06:20:02 UTC
*** Bug 1558306 has been marked as a duplicate of this bug. ***

Comment 9 Nicholas Nachefski 2018-03-20 13:45:39 UTC
The playbook is all over the place in it's current state.  In case anyone wants a reference for what works now (in 3.9), here you go...,${component}:${version}

Comment 11 Clayton Coleman 2018-03-26 03:31:59 UTC is a first step towards enforcing a consistent internal pattern based on oreg_url, other work will complete this by publishing images with consistent names.

Comment 12 Scott Dodson 2018-05-02 14:55:14 UTC
*** Bug 1573904 has been marked as a duplicate of this bug. ***

Comment 13 Scott Dodson 2018-05-08 13:56:41 UTC
*** Bug 1575994 has been marked as a duplicate of this bug. ***

Comment 14 Scott Dodson 2018-05-10 14:06:36 UTC
*** Bug 1576826 has been marked as a duplicate of this bug. ***

Comment 19 Johnny Liu 2018-05-29 10:41:44 UTC
Seem like the PR already merged into openshift-ansible-3.10.0-0.53.0, so change target release.

Comment 24 Michael Gugino 2018-05-30 13:56:57 UTC
PR for checking for removed variables:

etcd oreg_url support:

Comment 25 Nicholas Nachefski 2018-05-31 03:44:23 UTC
As of 3.10.0-0.54.0, i still needed these vars (along with oreg_url) to get a successful install.

Looks like just etcd and CNS stuff needs fixing.

Comment 26 Scott Dodson 2018-05-31 13:12:49 UTC can likely be resolved by addressing comment 25 as well.

Comment 27 Michael Gugino 2018-05-31 17:30:57 UTC
Glusterfs PR:

Comment 28 Nicholas Nachefski 2018-05-31 17:32:57 UTC
In addition to my previous comment, the registry web console appears to also not be using oreg_url.

Warning   Failed                        kubelet,     Failed to pull image "": rpc error: code = Unknown desc = All endpoints blocked.

I couldnt even get it to work the old way, by setting these:

So etcd, CNS, and registry-console seems to not be respecting oreg_url in 3.10.0-0.54.0.

Comment 29 Scott Dodson 2018-05-31 19:06:48 UTC
The fixes for etcd and openshift-web-console weren't in 3.10.0-0.54.0. They'll be in the next build tonight/tomorrow.

The only unmerged fix is GlusterFS right now and that should merge today.

Comment 30 Nicholas Nachefski 2018-06-01 14:51:41 UTC
Looks good now.

I got a clean (disconnected) install of 3.10.0-0.57.0 using only the oreg_url variable.  This will make it alot easier to do these kinds of deployments.  Thank you very much!


Comment 31 Scott Dodson 2018-06-01 14:58:10 UTC
In 3.10.0-0.57.0 and later.

Comment 33 Scott Dodson 2018-06-04 12:12:48 UTC reverts a recent change that should cause the problem you're describing.

Comment 34 Johnny Liu 2018-06-05 07:03:28 UTC
(In reply to Scott Dodson from comment #33)
> reverts a recent
> change that should cause the problem you're describing.

No, they are different issue, PR#8616 would fix BZ#1584966, but in this bug we are tracking oreg_url should be used in registry-console, no need to set any separate options (for registry console, the option name is "openshift_cockpit_deployer_prefix"), PR#8616 would not fix such issue, registry-console would still need specify openshift_cockpit_deployer_prefix to your local registry.

After go through the openshift-ansible code, regsitry-console would be deployed from *registry-console template* installed by openshift-ansible, openshift_cockpit_deployer_prefix is used to overwrite the default when running "oc new-app --template=registry-console" command, for such template or ImageStream template, installer also have another option - openshift_examples_modify_imagestreams to modify the default registry to oreg_url registry without need openshift_cockpit_deployer_prefix setting. 

So registry-console is not in the scope of this bug.

Comment 35 Johnny Liu 2018-06-05 07:04:29 UTC
Based on comment 32, move this bug to verified.

Comment 38 Scott Dodson 2018-06-21 15:09:38 UTC
*** Bug 1516535 has been marked as a duplicate of this bug. ***

Comment 40 Johnny Liu 2018-07-10 11:32:20 UTC
Copied comments from
> The following variables need to be kept
> openshift_cockpit_deployer_prefix
> openshift_cockpit_deployer_version
I think I already explained the reason of the two variables should be kept in

> openshift_provisioners_image_prefix
> openshift_provisioners_image_version
This is my first time to see this two vars, have some search in openshift-ansible code, seem like they are related to storage component, I am not familiar with that component. @jhou do you know that?

BTW, this bug is tracking too much components and vars, if  openshift_provisioners_image_* could be confirmed to a real issue, it is better to create another bug to track.

Comment 41 Kenjiro Nakayama 2018-07-22 06:53:30 UTC
*** Bug 1604714 has been marked as a duplicate of this bug. ***

Note You need to log in before you can comment on or make changes to this bug.