detect and handle CRD breaking changes

[a-hilaly]

Feature Description

Currently kro does not detect breaking changes in ResourceGroup schemas when users modify them. Breaking changes like adding new required fields or removing existing fields can cause instance reconciliation failures. we need to add support for detecting such changes and marking the ResourceGroup as invalid. The system should also support an annotation
(kro.run/allow-breaking-changes: "true") to explicitly allow these changes when needed.

Examples:

# Original ResourceGroup schema
apiVersion: kro.run/v1alpha1
kind: ResourceGroup
metadata:
  name: my-application
spec:
  schema:
    spec:
      name: string
      image: string | default="nginx"

# Breaking change - adding a required field without default value
apiVersion: kro.run/v1alpha1
kind: ResourceGroup
metadata:
  name: my-application
  annotations:
    kro.run/allow-breaking-changes: "true"  # Required to apply breaking changes
spec:
  schema:
    spec:
      name: string
      image: string | default="nginx"
      port: number | require=true  # Breaking change: new required field

[frewilhelm]

hey @a-hilaly

currently I am making myself familiar with CRDs created by kro/RGD and stumbled upon this issue. I got a question which problems this issue addresses:

If I understand it correctly, the purpose of this task is to only allow an update of an existing RGD that specifies a CRD with the same kind and apiVersion if (a) only optional fields or required fields with default values are added to the spec.schema.spec , or (b) the annotation kro.run/allow-breaking-changes: "true" is set in the RGD.

• Why do you want to solve this using an annotation on the RGD?
  I would argue that a required field in the spec.schema about controlling the behaviour could make more sense as users should definitely know how the RGD will behave on changes. I could imagine something like strategy: create (default), upgrade (on non-breaking changes), replace.
• What about resources that are changed in the ResourceGraphDefinition? They are not part of the resulting CRD but I could imagine that they become orphaned when an instance of the updated CRD/RGD is deleted.

WDYT?

[github-actions]

This issue has been marked stale due to 30 days of inactivity. Maintainers use this label for hygiene. Comment or add a relevant label (keepalive) to keep it from becoming stale.