Skip to content

Converge LLMModel reconciler onto the PassthroughModel gateway-apply failure handling #109

Description

@dcmcand

The PassthroughModel reconciler treats a failed AI-Gateway-resource apply (e.g. CRDs not yet installed) as an ApplyFailed condition + Error phase + one-minute requeue (passthroughmodel_controller.go:131). The older LLMModel reconciler logs-and-continues for the identical case (llmmodel_controller.go:445-470, "CRD may not be installed, skipping") with no Error phase and no requeue.

The surface-and-requeue behavior is the better of the two and is now documented in design.md as the intended operator-wide convention. This issue tracks converging the LLMModel path onto it so the operator handles the same situation one way.

Surfaced in review of #96.

Acceptance

  • LLMModel reconciler sets an ApplyFailed condition and requeues on a degraded gateway apply
  • Existing LLMModel tests updated
  • Behavior matches the convention documented in design.md

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Fields

    Priority

    None yet

    Start date

    None yet

    Target date

    None yet

    Size

    None yet

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions