Skip to main content

generator.yaml Configuration Reference

The generator.yaml file is the primary way you control how the code generator builds a service controller. It lives at the root of each service controller repository and tells the code generator which resources to generate, how to map AWS API fields to Kubernetes CRD fields, and where to inject custom behavior.

Every service controller has one. When you first bootstrap a controller, all resources are listed in ignore.resource_names. As you add support for each resource, you remove it from the ignore list and add its configuration.

Top-Level Structure​

ignore:
  resource_names:
    - ResourceToSkip
  field_paths:
    - CreateInputShape.FieldToSkip

resources:
  ResourceName:
    # Resource-level configuration (exceptions, hooks, fields, renames, tags, reconcile)

operations:
  OperationName:
    # Operation-level configuration (wrappers, custom methods)

The three top-level sections are:

  • ignore — What to skip during generation
  • resources — Per-resource configuration (this is where most of your config lives)
  • operations — Per-API-operation configuration (wrappers, custom output methods)

Ignore Configuration​

ignore.resource_names​

List resources that should not be generated. Used in bootstrap PRs to scaffold a controller without generating any resource-specific code. Remove a resource from this list when you're ready to add it.

ignore.field_paths​

Skip specific fields during code generation. Use the format ShapeName.FieldName to target fields in specific API operation shapes:

ignore:
  field_paths:
    - VersioningConfiguration.MFADelete
    - CreateUserInput.AuthenticationMode

Resource Configuration​

exceptions​

Configure how the controller handles AWS API errors.

404 error codes — Many AWS APIs return non-standard error codes for "not found":

resources:
  Table:
    exceptions:
      errors:
        404:
          code: ResourceNotFoundException

Terminal codes — Error codes that indicate the resource cannot be reconciled without user intervention:

resources:
  DBCluster:
    exceptions:
      terminal_codes:
        - DBClusterQuotaExceededFault
        - InvalidParameterValue
        - InvalidParameterCombination

reconcile​

Control reconciliation behavior:

resources:
  NotebookInstance:
    reconcile:
      requeue_on_success_seconds: 60

By default, successfully reconciled resources are requeued after 10 hours. Set requeue_on_success_seconds for resources whose state changes frequently outside of Kubernetes (e.g., via the AWS console).

is_arn_primary_key​

Set to true when a resource has no Name or ID field and can only be identified by ARN:

resources:
  ModelPackage:
    is_arn_primary_key: true

renames, fields, hooks, tags​

These are covered in their own pages:

Operation Configuration​

output_wrapper_field_path​

Some AWS APIs wrap responses in a nested object. Use this to flatten the wrapper so its fields land directly in Spec/Status:

operations:
  CreateReplicationGroup:
    output_wrapper_field_path: ReplicationGroup
  DescribeReplicationGroups:
    output_wrapper_field_path: ReplicationGroups  # plural for list operations

input_wrapper_field_path​

Some AWS APIs wrap input fields in a nested structure. This flattens the wrapper's fields into the CRD Spec:

operations:
  CreateBackupPlan:
    input_wrapper_field_path: BackupPlan
    output_wrapper_field_path: BackupPlan
Fields outside the wrapper

Fields on the API input that are outside the wrapper struct are excluded from the CRD spec. If those fields are required for CRUD operations, you need custom hooks to wire them in. See Hooks and Custom Code for details.

set_output_custom_method_name​

Specify a custom Go method to call when setting output fields from an API response:

operations:
  CreateUser:
    set_output_custom_method_name: CustomCreateUserSetOutput

custom_implementation​

Replace the entire generated SDK call with a custom implementation:

operations:
  ModifyUser:
    custom_implementation: CustomModifyUser

Next Steps​

Built with ♥ by AWS