Automated cherry pick of #76722: Pin GCE Windows node image to 1809 v20190312.

[pjh]

Cherry pick of #76722 on release-1.14.

#76722: Pin GCE Windows node image to 1809 v20190312.

[pjh]

/assign yujuhong

This will hopefully get https://testgrid.k8s.io/google-windows#windows-gce-1.14&width=20&show-stale-tests= back to green.

[yujuhong]

/lgtm
/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: pjh, yujuhong

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• cluster/gce/OWNERS [yujuhong]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[yujuhong]

/assign @tpepper @aleksandra-malinowska
for cherrypick approval

[yujuhong]

Windows GCE test job against 1.14 has been failing because of the bug in the new Windows image. We need to cherrypick the fix to unblock the tests: https://testgrid.k8s.io/google-windows#windows-gce-1.14

[pjh]

/retest

[tpepper]

It would have been better if the upstream commit message said more clearly "Docker 18.09" instead of the more opaque "1809" and included some of the details on what has been found wrong in newer patch versions. #76666 has details on what's happening here so there is some connection. But lacking details clearly in individual commits, and then obfuscated again through cherry-picking, this is a type of change somebody in the future is likely to re-change and cause a regression. Somebody in the future is likely to see old docker, need new docker for some reason, but they wont know why it was previously pinned or how to validate the version they want also includes the right fixes for the regressed behavior causing the original pinning.

I suppose it's too late to make this more clear since the cherry pick comes from something merged in master. But please try to capture this more clearly in future commits...

[yujuhong]

@tpepper I agree the release note can be more clear, but this is actually "Windows Server image 1809", and has nothing to do with the Docker version.

[yujuhong]

@tpepper I agree the release note can be more clear, but this is actually "Windows Server image 1809", and has nothing to do with the Docker version.

A buggy patch was introduced to the Windows image and broke the tests, and that's why we pinned the image to a version prior to the the OS patch.

[tpepper]

ACK

Thanks for the clarification. All the discussion of docker 18.09 versions in the issue report mislead me once the focus shifted to the similarly named windows node OS image.

[pjh]

Sorry for the confusion. If anyone has specific suggestions for how the commit message ("Pin GCE Windows node image to 1809 v20190312.") or release note ("Windows nodes on GCE use a known-working 1809 image rather than the latest 1809 image.") could have been better, or if there are guidelines for writing a good release note, please let me know.