terraform-provider-aws: Using the aws_serverlessapplicationrepository_cloudformation_stack resource works the first time, but subsequent updates fail.

So - interesting that there seems to be a second issue. For applications that require parameters (my real use case, not the hello-world one)… once i hard coded the CAPABILITY_IAM for this stack, I noticed a new bug. It seems that the parameters are not being persisted. Terraform refresh pulls back 4 parameters, even though there are 7 parameters in the configuration. I’ve not noticed anything strange about the parameters being dropped vs. kept, but something to look into. I’ll try to dig deeper but the result is that this component is still not working for me given this second issue.

I can confirm that this problem still exists in the latest version of the provider even after almost 2 year. As @cdhesse said, if you have more than 4 parameters, somehow the plan only refreshes up to 4 parameters from current stack that exists in AWS, and the rest of the parameters are treated as new, hence a diff is calculated. But when you apply the plan, since the diff that was calculated is not the reality, apply errors out with change set: unexpected state 'FAILED', wanted target 'CREATE_COMPLETE'. last error: No updates are to be performed error

Also seeing this issue.

We’ve deployed a stack using

resource "aws_serverlessapplicationrepository_cloudformation_stack" "salesforceIntegrationLambda" {
    name            = "salesforceIntegrationLambda"
    application_id  = "arn:aws:serverlessrepo:us-west-2:821825267871:applications/AmazonConnectSalesforceLambda"

    capabilities = [ 
        "CAPABILITY_IAM",
        "CAPABILITY_NAMED_IAM"
    ]

    parameters = {
      ...
   }
}

which creates a cloudformation stack called serverlessrepo-salesforceIntegrationLambda.

However, on the second terraform apply, we get a failure. Checking CloudTrail, I can see that a cloudformation.DescribeStacks API call is failing with the error Stack with id aws-serverless-repository-salesforceIntegrationLambda does not exist - it seems to be looking for the wrong stack name?

My temporary solution for this was to add a tag with a random string. this forces an update on every apply though, so it’s not ideal.

resource "aws_serverlessapplicationrepository_cloudformation_stack" "some_app" {
  tags = {
    RollMe = random_string.rollme.result
  }
}

resource "random_string" "rollme" {
  length  = 16
  special = false

  keepers = {
    uuid = "${uuid()}"
  }
}

My work around to this issue. just use the local .json or .yml CloudFormation template with terraform aws_cloudformationn_stack resource. Then terraform apply works fine without destroying the CloudFormation stack resource.