Description of problem:

A OSD cluster installation on the GCP cloud provider failed to provision worker machines successfully.

Analysis of the machine-api-controller logs showed that it was unable to resolve oauth2.googleapis.com. This occurred at 2022-10-10 23:56

The dns-operator logs show that it did not begin to install until the same time, and was in the process of installing whilst the machine-api-controller was failing to DNS resolve the GCP API.

The machine-api-controller gave up on provisioning the failed workers before the DNS operator seemed to become ready to allow it to resolve DNS.

I was able to later confirm that DNS resolution was fine in the machine-api-controller after the DNS operator installed, by deleting the machine that failed to provision, and confirming that a replacement machine could successfully provision.

Version-Release number of selected component (if applicable):

4.11.7

Expected results:

Cluster DNS should be in an installed state before machine-api pods should be attempting to resolve external DNS addresses.