Feature Overview

Where is the infrastructure? An important question that determines the behavior of the OpenShift cluster operator with HyperShift. First, let us try to understand what is meant by "infrastructure" here. Infrastructure is any resource definitions that need to exist before attempting to bootstrap an OpenShift / Kubernetes cluster, this can range from Virtual Private Networks (VPCs), Security groups, firewall rules, additional day-0 load balancers, etc.

Now that we understand what is meant by Infrastructure, let's answer the "where" part. Some customers prefer managing these infra resources themselves, i.e., no permissions are to be given to any tooling for infrastructure automation. Other customers prefer full automation.

This feature is about designing for both scenarios.

Background, and strategic fit

It is important to give customers choice on how they want their infrastructure to be managed.

Goal(s)

• Consider both BYO and automated infrastructure models in the design of HyperShift. This will likely reflect on API definitions (e.g., existing VPC vs one to be created), as well as introduce different code paths to cover both cases.
• Identify integration points with other offering to cover both models (e.g., AI)
  • Machineset -> Machine -> {AWS/GCP/Azure}Machine <- CAPI actuator
  • Machineset -> Machine -> AgentDrivenHostInstallation <- Assisted Installer Operator
  • Machineset -> Machine -> AgentDrivenHostInstallation <- Assisted Installer Operator <- CAPI actuator

This feature can be related to the consumption models feature [1]. 

References

[1] https://issues.redhat.com/browse/OCPPLAN-5662