Background

For clusters upgraded to 4.13 or above, GCP is supported by the CPMS but we do not automatically create the CPMS for them.

We should generate the CPMS as Inactive in the same way we do for AWS.

The following should be done:

• Get the provider config for the newest machine
• Remove the failure domain fields from it (zone)
• Fetch the failure domains from the machines
• Fetch the failure domains from the machinesets
• Union the two failure domain sets

Steps

• Review the generator code
• Identify the integration points where it becomes cloud specific
• Build the GCP specific integration
• Add tests for generating on GCP

Stakeholders

• Cluster Infra

Definition of Done

• On GCP clusters, a generated CPMS is created and is set to Inactive

• Docs

• Update in repo docs that talk about the CPMS installation to reflect these changes

• Testing

• Build an GCP cluster, check the CPMS looks healthy (eg replicas are updated)