Fix bug in backend syncer where backend service was being created without health check

[rramkumar1]

Introduced by #424. Looks like we can't create a BackendService without a health check. Fixed the BackendPool.Create implementation to now take in a health check link.

/assign @bowei

[bowei]

Did this get detected in our e2e tests?

[bowei]

I remember you described the new flow as updating the health check later. Is this going to be updating the healthcheck an additional time?

[rramkumar1]

Yeah, e2e's detected this.

See: https://k8s-gubernator.appspot.com/build/kubernetes-jenkins/logs/ci-ingress-gce-e2e/2694

I'm not sure why the test hasn't timed out yet, but if you look at the logs, you can see ingresses failing to acquire IP's. Further investigation of GLBC logs clearly indicated the issue.

[rramkumar1]

No this will go back to the previous behavior where we create the backend service w/ a HC link and then do an "ensure" on the health check link. The ensure of the health check link right after a create will obviously do nothing since the health check link should be up-to-date already.

When a backend is updated, the health check could change. In any case, we should still only sync a health check once per backend sync.

[bowei]

/lgtm