Bug: References always converted into strings for Step Functions definition substitutions - Can't use a CF parameter for TimeoutSeconds attribute (Integer)

[alexcasalboni]

Description: when I create a AWS::Serverless::StateMachine resource, I'd like to parametrize the value of my state's TimeoutSeconds attribute using a CloudFormation parameter.

I've defined the parameter as follows:

Parameters: 
  myParameter:
    Type: Number
    Default: 300

and I'm referencing it in my task definition as follows:

myTask:
  Type: Task
  Resource: !GetAtt myfunction.Arn
  End: true
  TimeoutSeconds: !Ref myParameter

Steps to reproduce the issue:

1. Define a CloudFormation parameter
2. Define a Step Functions state
3. Try to reference the parameter in the TimeoutSeconds attribute

Observed result:

The deployment fails with this error message:

Invalid State Machine Definition: 'SCHEMA_VALIDATION_FAILED: Expected value of type Integer

Expected result:

I'd expect the deploy to succeed since the parameter type and value are valid.

My understanding is that CloudFormation treats every parameter as a string under the hood, even if you define it as Type: Number. But then how do I get to use this parameter as an Integer, as SAM is expecting? Is this a SAM issue or a Step Functions issue?

[alexcasalboni]

Update: I have inspected the changeset to verify how SAM is transforming the state machine definition.

This is the transformed definition (just the snippet corresponding to my task):

"            \"Iterator\": {",
"                \"StartAt\": \"Executor\",",
"                \"States\": {",
"                    \"Executor\": {",
"                        \"End\": true,",
"                        \"Resource\": \"${definition_substitution_2}\",",
"                        \"TimeoutSeconds\": \"${definition_substitution_3}\",",
"                        \"Type\": \"Task\"",
"                    }",
"                }",
"            },",

As you can see, the TimeoutSeconds attribute is being transformed into a string: \"TimeoutSeconds\": \"${definition_substitution_3}\",".

So I think there is something wrong with the SAM transformation, probably somewhere around here: https://github.com/awslabs/serverless-application-model/blob/develop/samtranslator/model/stepfunctions/generators.py#L254

[bradduhon]

Glad this is being addressed already!

I just came over here to also add this is an issue when using Fn::FindInMap for instance when I have different TimeoutSeconds between dev, test & prod.

[bradduhon]

Not sure if this is the place to post this or open a new report.

This is also an issue with the retry fields: MaxAttempts & BackoffRate. I imagine it is also an issue in the IntervalSeconds but I have not yet tested that...

MaxAttempts: !FindInMap: [Integration, !Ref STAGE, statesMaxAttempts]
BackoffRate: !FindInMap: [Integration, !Ref STAGE, statesBackoffRate]

Will convert the Integer stored in the Map to a String and throw an error on deployment. This worked in the normal Non-SAM StepFunctions resources.

[bradduhon]

Is this still being worked on the branch doesn't seem to have seen activity for a month now.

[piersf]

This seems to still be a problem. We basically can't pass any integer values using the DefinitionSubstitutions property to the state machine definition template(in JSON format) residing in a different file. The only solution I've found so far is to hardcode in the state machine definition template all integer values.

[mattshep]

I encountered this issue today when trying to use the value of a parameter for the MaxAttempts attribute of a Retry block.

Template Snippets:

  FetchMaxRetryAttempts:
    Type: Number
    Description: "Maximum number of retry attempts for Fetch stage (default: 3)"
    Default: 3
    MinValue: 0
    MaxValue: 3

            Retry:
              -
                ErrorEquals:
                  - "States.ALL"
                IntervalSeconds: 600
                MaxAttempts: !Ref FetchMaxRetryAttempts

Error:

Invalid State Machine Definition: 'SCHEMA_VALIDATION_FAILED: Expected value of type Integer at /States/Fetch/Retry[0]/MaxAttempts' (Service: AWSStepFunctions; Status Code: 400; Error Code: InvalidDefinition; Request ID: 2c1fb4e0-5a62-432f-91a7-ff5e225dd903; Proxy: null)

[jtheuer]

Should this be fixed for TargetSpotCapacity as well? I still cannot use a substitution here.

[mikebroberts]

I just hit this for TimeoutSeconds too. Any plan on a fix for this?