Background

Not every platform needs to support failure domains, for example baremetal.

In order to support generic platforms, we could introduce a generic providerConfig abstraction that can handle the case when the failureDomains is empty.

In this case, we never manipulate the providerSpec anyway so there's no need to have a platform specific providerConfig abstraction.

Steps

• Create a generic providerConfig abstraction
• Write tests for the generic providerConfig abstraction
• Integrate this so that we use the generic abstraction when there are no failure domains
• Test on a platform that doesn't support failure domains, eg vSphere

Stakeholders

• Cluster Infra
• Platforms who aren't AWS, GCP, Azure

Definition of Done

• We can support generic platforms

• Docs

• Announce single zone support for any platform

• Testing

• Manual testing against single zone and other platforms such as vSphere