Do not merge. Will re-cut after #8833 is in to fix gke-e2e runs.

cc @thockin @quinton-hoole re. other e2e failures.

In addition to outstanding e2e failures, we have a backlog of 28 PRs with lgtm, which include the last v1beta1-to-v1beta3 conversions, the last v1 service API change, and some significant kubectl command (run, scale, exec) changes. I'd like to postpone the 0.18 cut until we get these in. In particular, we want to have a release with full v1beta3 support before we rip out v1beta1 and v1beta2. 0.18 was expected to be that release, but the merge freeze interfered with that plan.

cc @caesarxuchao @krousey

e2e is stable again. re-cut release from 17f6f86. PTAL

LGTM.

@bgrant0607 I believe all of the v1beta3 PRs have been merged prior to the v0.18.0 tag.

cc @vishh
gke-ci is still a bit flaky (it was broken overnight by #8740), but I'd still like to get this in today so we can get more test coverage over the weekend.

Yes, thanks. Please cut the release. v1beta1 and v1beta2 will be ripped out first thing Monday morning.

What was the gke-ci breakage due to, specifically? The env. var. change, flag change, salt change, ...?

The envvar change broke gke-ci. Robby fixed this morning.

Can has merge?