[Bug]: aws_cloudwatch_metric_alarm - first apply fails, second succeeds - issue with metric_query block and the period parameter

Terraform Core Version

1.6.4

AWS Provider Version

5.16.1

Affected Resource(s)

aws_cloudwatch_metric_alarm

Expected Behavior

Apply should work first time.

Actual Behavior

First apply fails and then a second apply works.

Relevant Error/Panic Output Snippet

No response

Terraform Configuration Files

Note the period = line and the comments

resource "aws_cloudwatch_metric_alarm" "cw_ec2_statuscheck_nat_euw1a" {
  provider          = aws.ou
  alarm_name        = "${local.name}-EC2-StatusCheck-NAT-euw1a"
  alarm_description = "Checks status of ${local.name}-NAT-euw1a."

  dynamic "metric_query" {
    for_each = toset(data.aws_instances.nat_instance_euw1a.ids)
    content {
      id = replace(metric_query.value, "-", "")
      metric {
        namespace   = "AWS/EC2"
        metric_name = "StatusCheckFailed"
        dimensions  = { InstanceId = metric_query.value }
        stat        = "Average"
        period      = 300
      }
    }
  }

  metric_query {
    id          = "e1"
    expression  = "AVG(METRICS())"
    label       = "StatusCheck"
    return_data = "true"
    ##### period = 0 ##### tried with no period and period = 0.  Apply always fails first time and works the second time
  }

  evaluation_periods  = "1"
  comparison_operator = "GreaterThanOrEqualToThreshold"
  threshold           = var.threshold_ec2_status_check

  alarm_actions = [aws_sns_topic.sns_topic.arn]
}

Steps to Reproduce

Setup config as above, obviously needs some refactoring to work in another environment, and run an apply.

Debug Output

After the first apply the error is:

│ Error: creating CloudWatch Metric Alarm (tcccomniverse-dev-EC2-StatusCheck-NAT-euw1a): ValidationError: Period must not be null
│       status code: 400, request id: 32df7504-99bf-41e5-a0a5-0ef6afd8f0d4
│
│   with module.terraform-module-environment.aws_cloudwatch_metric_alarm.cw_ec2_statuscheck_nat_euw1a,
│   on modules/main/cloudwatch.tf line 105, in resource "aws_cloudwatch_metric_alarm" "cw_ec2_statuscheck_nat_euw1a":
│  105: resource "aws_cloudwatch_metric_alarm" "cw_ec2_statuscheck_nat_euw1a" {

On the second apply it is successful.

Panic Output

No response

Important Factoids

The period parameter when used in a metric_query block should be optional.

It seems it is optional but that on a first apply the provider is confused, and always shows the errors if the period is set to 0 or omitted.

This happens if period = 0 is set or if period is removed altogether.

I will setup a simpler example as soon as possible as the above is code taken directly from my current deployed code.

References

No response

Would you like to implement a fix?

None

hashicorp / terraform-provider-aws