Broker ansible locks up determining if local_openshift_enabled

[eriknelson]

So it looks like under certain circumstances, the broker's ansible will completely lock up and hang when dynamically trying to determine if the local openshift adapter is enabled. Currently it uses the following line set in the defaults file:

broker_local_openshift_enabled: "{{ ('image.openshift.io' in lookup('k8s', cluster_info='api_groups')) | bool }}"

This task will hang as it tries to build the config: https://github.com/openshift/ansible-service-broker/blob/master/ansible_role/tasks/build_config.yml#L19

It's not clear to me exactly why or when this shows up, because I've seen it work in some circumstances and not in others. Right now, in the current OLM setup with catasb, you can't deploy a broker due to this issue unless you override the var with broker_local_openshift_enabled: true.

Here are the operator logs: https://gist.github.com/eriknelson/0b6d82a2de646374dd565f84ccbb9824

[jmontleon]

May be related to: ansible/ansible-runner#154

Possible workaround: https://github.com/ansibleplaybookbundle/apb-base/pull/52/files

[jmontleon]

After some testing this is most like due to kubernetes-client/python#653

We can pin the versions of python openshift and kubernetes until the underlying problem in python kubernetes is fixed. water-hole/ansible-operator#56

[jmontleon]

Appears to be fixed in kubernetes 8.0.0 and openshift 0.8.0, which are both now released.

[fabianvf]

No that label seems appropriate

[djzager]

/shrug

[jmrodri]

Downstream bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1649078

[jmontleon]

kubernetes 8.0.0 and openshift 0.8.0 have been built upstream and downstream and look to resolve the issue.