Terraform regenerating resources when a new object is added to map - amazon-web-services

Question

I have a Terraform (pre 0.12) module that generates an Amazon Cognito user pool + a client and domain.
resource "aws_cognito_user_pool" "pool" {
count = "${var.user_pool_count}"
name = "${lookup(var.user_pools[count.index], "name")}"
username_attributes = ["email"]
auto_verified_attributes = ["email"]
password_policy {
minimum_length = "${lookup(var.user_pools[count.index], "password_minimum_length")}"
require_lowercase = "${lookup(var.user_pools[count.index], "password_require_lowercase")}"
require_numbers = "${lookup(var.user_pools[count.index], "password_require_numbers")}"
require_symbols = "${lookup(var.user_pools[count.index], "password_require_symbols")}"
require_uppercase = "${lookup(var.user_pools[count.index], "password_require_uppercase")}"
}
verification_message_template = {
default_email_option = "CONFIRM_WITH_LINK"
}
lambda_config = {
pre_token_generation = "${var.lambda_pre_token_generation}"
custom_message = "${var.lambda_custom_message}"
}
email_configuration = {
reply_to_email_address = "${lookup(var.user_pools[count.index], "reply_to_email_address")}"
source_arn = "${lookup(var.user_pools[count.index], "source_arn")}"
email_sending_account = "${lookup(var.user_pools[count.index], "email_sending_account")}"
}
schema = [
< REDACTED >
]
}
resource "aws_cognito_user_pool_client" "client" {
count = "${var.user_pool_count}"
name = "${lookup(var.user_pools[count.index], "name")}"
user_pool_id = "${element(aws_cognito_user_pool.pool.*.id,count.index)}"
explicit_auth_flows = ["ADMIN_NO_SRP_AUTH", "USER_PASSWORD_AUTH"]
}
resource "aws_cognito_user_pool_domain" "main" {
count = "${var.user_pool_count}"
domain = "${lookup(var.user_pools[count.index], "domain")}"
user_pool_id = "${element(aws_cognito_user_pool.pool.*.id,count.index)}"
}
This accepts a list of maps called user_pools to define the Cognito user pools required.
Unfortunately, when I add a new map with the definition of a new pool in it, Terraform forces the recreation of aws_cognito_user_pool_client and aws_cognito_user_pool_domain for all pools. This appears to be because it sees a change in:
user_pool_id: "eu-west-1_R8SDX8Yqj" => "${element(aws_cognito_user_pool.pool.*.id,count.index)}" (forces new resource)
I am assuming this is because Terraform is seeing a change in aws_cognito_user_pool.pool.*.id and forcing the recreation. Can anyone explain how to get around this? It is suboptimal for me to have all of my domains and clients regenerated.

Answer 1

For anyone reading this. I found the following issue on Github - https://github.com/hashicorp/terraform/issues/14357
Changing my syntax to the following appeared to fix it.
user_pool_id = "${aws_cognito_user_pool.pool.*.id[count.index]}"