[Bug]: r/aws_cloudformation_stack_set_instance: Resource deletion ignores operation_preferences

Terraform Core Version

1.9.4

AWS Provider Version

5.62.0

Affected Resource(s)

aws_cloudformation_stack_set_instance

Expected Behavior

During resource creation (resourceStackSetInstanceCreate()), the configuration's operation_preferences should be passed to the CreateStackInstances API operation, e.g. targets should be created according to any specified concurrency limits.
During resource update (resourceStackSetInstanceUpdate()), the configuration's operation_preferences should be passed to the UpdateStackInstances API operation, e.g. targets should be updated according to any specified concurrency limits.
During resource deletion (resourceStackSetInstanceDelete()), the configuration's operation_preferences should be passed to the DeleteStackInstances API operation, e.g. targets should be deleted according to any specified concurrency limits.

Actual Behavior

[EXPECTED] During resource creation (resourceStackSetInstanceCreate()), the configuration's operation_preferences are passed to the CreateStackInstances API operation, e.g. targets are created according to any specified concurrency limits.
[EXPECTED] During resource update (resourceStackSetInstanceUpdate()), the configuration's operation_preferences are passed to the UpdateStackInstances API operation, e.g. targets are updated according to any specified concurrency limits.
[UNEXPECTED] During resource deletion (resourceStackSetInstanceDelete()), the configuration's operation_preferences are not passed to the DeleteStackInstances API operation, e.g. targets are not deleted according to any specified concurrency limits, instead they are always deleted sequentially (the default MaxConcurrentCount value of 1).

Relevant Error/Panic Output Snippet

No response

Terraform Configuration Files

resource "aws_cloudformation_stack_set_instance" "this" {
  stack_set_name = "MyStackSet"

  ...

  operation_preferences {
    max_concurrent_percentage = 100
  }
}

Steps to Reproduce

terraform apply -auto-approve (quick, because concurrency is configured to 100%)
terraform destroy -auto-approve (slow, because concurrency is the default value of 1)

Alternatively, this can be reproduced using terraform test as appropriate.

Debug Output

No response

Panic Output

No response

Important Factoids

Just a thought: theoretically, there could be a scenario where separate operation_preferences configurations are required between each of the resource create/update/delete operations, e.g. have separate create_operation_preferences, update_operation_preferences and delete_operation_preferences. However, this is probably a rare scenario and the additional implementation complexity would be overkill.

References

https://github.com/hashicorp/terraform-provider-aws/issues/33170#issuecomment-2291178463
Source code for resourceStackSetInstanceDelete(): https://github.com/hashicorp/terraform-provider-aws/blob/main/internal/service/cloudformation/stack_set_instance.go#L444C6-L444C36

Would you like to implement a fix?

Yes

hashicorp / terraform-provider-aws