do ugly things to wire a cloudconfig file #14444
Conversation
|
[test] No tests for this, so it probably works. |
| // note: we are passing a combined quota registry here... | ||
| kubePluginInitializer := kadmission.NewPluginInitializer(privilegedLoopbackKubeClientsetInternal, internalkubeInformerFactory, nil, nil, quotaRegistry) | ||
| kubePluginInitializer := kadmission.NewPluginInitializer(privilegedLoopbackKubeClientsetInternal, internalkubeInformerFactory, authorizer, cloudConfig, quotaRegistry) | ||
| originAdmission, kubeAdmission, err := buildAdmissionChains(options, privilegedLoopbackKubeClientsetInternal, pluginInitializer, kubePluginInitializer) |
There was a problem hiding this comment.
Why are we passing authorizer here now?
|
I have tested this to work on a multizone GCE cluster. Also added a card for writing a e2e test - https://trello.com/c/Ul2Ldjjn/508-write-an-e2e-test-for-multizone-fix |
|
nevermind, that was the controller refactor |
deads2k
force-pushed
the
api-03-cloduconfig
branch
from
de30259 to
ee1db3b
Compare
|
Evaluated for origin test up to ee1db3b |
|
continuous-integration/openshift-jenkins/test SUCCESS (https://ci.openshift.redhat.com/jenkins/job/test_pull_request_origin/1906/) (Base Commit: 5f2f3f4) |
Well, I thought we had a controller problem, not an admission one. I think we need to refactor the upstream admission plugin to use runtime information instead of depending on files on disk that deep into the init structure, but that would be 1.8 I think. |
|
There was a request to merge before the sprint opened, so I'll take that as lgtm. Doing it now [merge] |
|
Turns out we already have an e2e test for statically provisioned PVs - kubernetes/kubernetes#42829 . Although, it was added kind of recently. |
|
Evaluated for origin merge up to ee1db3b |
|
continuous-integration/openshift-jenkins/merge SUCCESS (https://ci.openshift.redhat.com/jenkins/job/merge_pull_request_origin/918/) (Base Commit: 748b1ab) (Image: devenv-rhel7_6327) |
Might fix https://bugzilla.redhat.com/show_bug.cgi?id=1454601
This punches through a couple layers to build an options object twice in order to plumb a value through an admission wiring step before admission can logically be completed to hopefully make an admission plugin work.
@eparis