fix: account-quota drift reduced

modules/account-quotas/README.md

@@ -38,9 +38,21 @@ aws --region us-east-1 service-quotas list-service-quotas --service-code ec2
 If you make a request to raise a quota, the output will show the requested value as `value` while the request is
 pending.
 
+### Special usage Notes
+
 Even though the Terraform will submit the support request, you may need to follow up with AWS support to get the request
 approved, via the AWS console or email.
 
+#### Resources are destroyed on change
+
+Because the AWS API often returns default values rather than configured or applicable values for a given quota, we have
+to ignore the value returned by the API or else face perpetual drift. To allow us to change the value in the future,
+even though we are ignoring it, we encode the value in the resource key, so that a change of value will result in a new
+resource being created and the old one being destroyed. Destroying the old resource has no actual effect (it does not
+even close an open request), so it is safe to do.
+
+### Example
+
 Here's an example snippet for how to use this component.
 
 ```yaml
@@ -128,5 +140,10 @@ components:
 - AWS CLI
   [command to list service codes](https://awscli.amazonaws.com/v2/documentation/api/latest/reference/service-quotas/list-services.html):
   `aws service-quotas list-services`
+- AWS CLI
+  [command to list service quotas](https://awscli.amazonaws.com/v2/documentation/api/latest/reference/service-quotas/list-service-quotas.html)
+  `aws service-quotas list-service-quotas`. Note where it says "For some quotas, only the default values are available."
+- [Medium article](https://medium.com/@jsonk/the-limit-does-not-exist-hidden-visibility-of-aws-service-limits-4b786f846bc0)
+  explaining how many AWS service limits are not available.
 
 [<img src="https://cloudposse.com/logo-300x69.svg" height="32" align="right"/>](https://cpco.io/component)

modules/account-quotas/main.tf

@@ -21,6 +21,26 @@ locals {
     quota_code   = quota.quota_code != null ? quota.quota_code : data.aws_servicequotas_service_quota.by_name[k].quota_code
     value        = quota.value
   } }
+
+  # Because the API often returns default values rather than configured or applicable values,
+  # we have to ignore the value returned by the API or else face perpetual drift.
+  # To allow us to change the value in the future, even though we are ignoring it,
+  # we encode the value in the resource key, so that a change of value will
+  # result in a new resource being created and the old one being destroyed.
+  # Destroying the old resource has no actual effect, it does not even close
+  # an open request, so it is safe to do.
+
+  quota_requests = { for k, quota in local.quotas_coded_map :
+    format("%v/%v/%v", quota.service_code, quota.quota_code, quota.value) => merge(
+      quota, { input_map_key = k }
+    )
+  }
+
+  quota_results = { for k, v in local.quota_requests : v.input_map_key => merge(
+    { for k, v in aws_servicequotas_service_quota.this[k] : k => v if k != "value" },
+    { "value reported (may be inaccurate)" = aws_servicequotas_service_quota.this[k].value },
+    { "value requested" = v.value }
+  ) }
 }
 
 data "aws_servicequotas_service" "by_name" {
@@ -37,9 +57,15 @@ data "aws_servicequotas_service_quota" "by_name" {
 }
 
 resource "aws_servicequotas_service_quota" "this" {
-  for_each = local.quotas_coded_map
+  for_each = local.quota_requests
 
   quota_code   = each.value.quota_code
   service_code = each.value.service_code
   value        = each.value.value
+
+  lifecycle {
+    # Literally about 50% of the time, the actual value set is not available,
+    # so the default value is reported instead, resulting in permanent drift.
+    ignore_changes = [value]
+  }
 }

modules/account-quotas/outputs.tf

@@ -1,4 +1,4 @@
 output "quotas" {
-  value       = aws_servicequotas_service_quota.this
+  value       = local.quota_results
   description = "Full report on all service quotas managed by this component."
 }