sagemaker: The YAML template for the PipelineDefinitionBody property is in Upper Camel Case (Pascal Case) while the template generated by the CDK has this property in Camel Case

[JawadAr]

Describe the bug

The YAML template for the AWS::SageMaker::Pipeline resource in the docs is as follows:

Type: AWS::SageMaker::Pipeline
Properties: 
  ParallelismConfiguration: 
    ParallelismConfiguration
  PipelineDefinition: 
    PipelineDefinition
  PipelineDescription: String
  PipelineDisplayName: String
  PipelineName: String
  RoleArn: String
  Tags: 
    - Tag

The PipelineDefinition property is further defined as follows:

  PipelineDefinitionBody: String
  PipelineDefinitionS3Location: 
    S3Location

Which means that a valid template could be:

Type: AWS::SageMaker::Pipeline
Properties: 
  PipelineDefinition: 
      PipelineDefinitionBody: String
  PipelineDescription: String
  PipelineDisplayName: String
  PipelineName: String
  RoleArn: String
  Tags: 
    - Tag

Notice that the property PipelineDefinitionBody is in Pascal Case. But what the CDK generates is:

Type: AWS::SageMaker::Pipeline
Properties: 
  PipelineDefinition: 
      pipelineDefinitionBody: String
  PipelineDescription: String
  PipelineDisplayName: String
  PipelineName: String
  RoleArn: String
  Tags: 
    - Tag

Notice that the property pipelineDefinitionBody is in Camel Case. Which is causing the following error in CloudFormation:

Properties validation failed for resource SampleCfnPipeline with message: #/PipelineDefinition: #: only 1 subschema matches out of 2 #/PipelineDefinition: #: 0 subschemas matched instead of one #/PipelineDefinition: required key [PipelineDefinitionBody] not found #/PipelineDefinition: extraneous key [pipelineDefinitionBody] is not permitted #/PipelineDefinition: required key [PipelineDefinitionS3Location] not found #/PipelineDefinition: extraneous key [pipelineDefinitionBody] is not permitted

The Python stack code for creating the template is as follows:

class MySageMakerStack(Stack):

    def __init__(self, scope: Construct, construct_id: str, **kwargs) -> None:
        super().__init__(scope, construct_id, **kwargs)
        
        pipeline_def = sagemaker.CfnPipeline.PipelineDefinitionProperty(
            pipeline_definition_body = "String"
        )

        cfn_pipeline = sagemaker.CfnPipeline(self, "SampleCfnPipeline",
            pipeline_definition=pipeline_def,
            pipeline_name="String",
            role_arn="String",
            pipeline_description="String",
            pipeline_display_name="String"
        )

Any generic code would reproduce the error.

Python CDK version: 2.81.0
CDK ToolKit version: 2.81.0
Machine: Macbook Pro 2020, 13-inch
OS: Ventura 13.0.1
IDE: VSCode

Expected Behavior

I expected the CDK to generate a valid template for the AWS::SageMaker::Pipeline resource. Which is as follows:

Type: AWS::SageMaker::Pipeline
Properties: 
  PipelineDefinition: 
      PipelineDefinitionBody: String
  PipelineDescription: String
  PipelineDisplayName: String
  PipelineName: String
  RoleArn: String
  Tags: 
    - Tag

Notice that the property PipelineDefinitionBody is in Pascal Case.

Current Behavior

What the CDK currently generates is:

Type: AWS::SageMaker::Pipeline
Properties: 
  PipelineDefinition: 
      pipelineDefinitionBody: String
  PipelineDescription: String
  PipelineDisplayName: String
  PipelineName: String
  RoleArn: String
  Tags: 
    - Tag

Notice that the property pipelineDefinitionBody is in Camel Case.

Reproduction Steps

Use the following stack in python CDK:

class MySageMakerStack(Stack):

    def __init__(self, scope: Construct, construct_id: str, **kwargs) -> None:
        super().__init__(scope, construct_id, **kwargs)
        
        pipeline_def = sagemaker.CfnPipeline.PipelineDefinitionProperty(
            pipeline_definition_body = "String"
        )

        cfn_pipeline = sagemaker.CfnPipeline(self, "SampleCfnPipeline",
            pipeline_definition=pipeline_def,
            pipeline_name="String",
            role_arn="String",
            pipeline_description="String",
            pipeline_display_name="String"
        )

Possible Solution

Possibly a typo error in converting from code to YAML.

Additional Information/Context

Python CDK version: 2.81.0
CDK ToolKit version: 2.81.0
Machine: Macbook Pro 2020, 13-inch
OS: Ventura 13.0.1
IDE: VSCode

CDK CLI Version

2.81.0

Framework Version

2.81.0

Node.js Version

18.16.0

OS

Ventura 13.0.1

Language

Python

Language Version

3.9.7

Other information

No response

[pahud]

Yes I can reproduce it in python but not in TypeScript. This seems to be a bug.

I'll discuss this with the team.

[pahud]

Hi @JawadAr

This is a cfnspec patch issue and we are tracking at #21767

At this moment, you will need to write your CDK in python like this:

sagemaker.CfnPipeline(self, "IssueTriagePyPipeline",
                              pipeline_name="IssueTriagePyPipeline",
                              role_arn="0000000000000000000000000:role/service-role/AmazonSageMaker-ExecutionRole-20210802T151519",
                              pipeline_definition={
                                  "PipelineDefinitionBody": "foo"
                              }
        )

And you should get the synthed resource like

Resources:
  IssueTriagePyPipeline:
    Type: AWS::SageMaker::Pipeline
    Properties:
      PipelineDefinition:
        PipelineDefinitionBody: foo
      PipelineName: IssueTriagePyPipeline
      RoleArn: 0000000000000000000000000:role/service-role/AmazonSageMaker-ExecutionRole-20210802T151519

Let me know if it works for you.

[JawadAr]

Hi @pahud
Yes, I figured this out today. This works, but the best practice is to provide a subclass from that L1 construct, which I am sure will be patched up soon. Thanks! 😊

[pahud]

closing this issue in favor of the tracking issue at #21767

[github-actions]

⚠️COMMENT VISIBILITY WARNING⚠️

Comments on closed issues are hard for our team to see.
If you need more assistance, please either tag a team member or open a new issue that references this one.
If you wish to keep having a conversation with other community members under this issue feel free to do so.