Terraform 在资源中创建一种或另一种块类型

Question

I am trying to create an alert policy within GCP using Terraform and Based on a boolean value I want to choose wether to create condition_threshold or conditionAbsent type policy. Since these are both different kinds of blocks within google_monitoring_alert_policy, I am not able to find a way to create either one or other based on a boolean value.

resource "google_monitoring_alert_policy" "logging_alert_policy" {
  for_each     = local.metrics_map
  project      = var.project
  display_name = lower(format("%s%s%s%s", join("_", [element(split("-", each.value.appcode), length(split("-", each.value.appcode)) - 1), element(split("-", each.value.appcode), 0)]), " ", "log_match ", each.value.display_name))

  combiner = each.value.how_to_trigger

  dynamic "conditions" {
    for_each = each.value.conditions

    content {
      display_name = lower(format("%s%s%s%s", join("_", [element(split("-", conditions.value.appcode), length(split("-", conditions.value.appcode)) - 1), element(split("-", conditions.value.appcode), 0)]), " ", "log_match ", conditions.value.display_name))

       

      # condition_threshold {

      #   filter = lower("metric.type=\"logging.googleapis.com/user/${format("%s%s%s%s", conditions.value.appcode, "-", lower(replace(lookup(conditions.value, "metric_name"), "/\\W|_|\\s/", "-")), "")}\" resource.type=\"${conditions.value.resource_type}\"")

      #   comparison      = conditions.value.comparison
      #   threshold_value = conditions.value.threshold_value
      #   duration        = conditions.value.duration

      #   trigger {
      #     count = lookup(conditions.value, "trigger_count", var.default_metric_alert["trigger_count"])
      #   }
      #   aggregations {
      #     alignment_period     = conditions.value.alignment_period
      #     per_series_aligner   = conditions.value.per_series_aligner
      #     cross_series_reducer = conditions.value.cross_series_reducer
      #     group_by_fields      = conditions.value.group_by_fields
      #   }
      # }

      condition_absent {
        
        aggregations {
          alignment_period     = "300s"
          per_series_aligner   = "ALIGN_MAX"
          cross_series_reducer = "REDUCE_MAX"
          group_by_fields      = ["resource.label.container_name"]
        }

        trigger {
          count = 1
        }

        duration        = "180s"

        filter = lower("metric.type=\"logging.googleapis.com/user/${format("%s%s%s%s", conditions.value.appcode, "-", lower(replace(lookup(conditions.value, "metric_name"), "/\\W|_|\\s/", "-")), "")}\" resource.type=\"${conditions.value.resource_type}\"")

      }

      
    }
  }

Answer 1

I would propose nested dynamic blocks that rely on local computed set of one element - or no elements.

Short example of that idea:

locals {
  absent = false  # let's say we want to have "condition_threshold" block if this is true 

  conditions_config        = local.absent == false ? ["1"] : []
  absent_conditions_config = local.absent == true ? ["1"] : []
 
  # I agree the above code is ugly but shows exactly the idea
}

# [...]

  dynamic "conditions" {
    for_each = each.value.conditions

    content {
      #[...]
       
      dynamic "condition_threshold" {
        for_each = toset(local.conditions_config)  # this will be one element if local.absent = false

        # content { ... } 
      }

      dynamic "condition_absent" {
        for_each = toset(local.absent_conditions_config) # this will be one element if local.absent = true

        # content { ... } 
      }