Description of problem:
Test PodNodeSelector on ocp-3.5, still have the same issue with upstream, using this bug to trace upstream PodNodeSelector related issue, just in case we forget it.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
I have PRs in flight to address these issues in kube upstream:
Reducing priority as this is not a release blocker.
Will review related PRs though.
Discussion in upstream PRs indicates that we're not likely to implement this.