RDS parameter group keeps being modified, even though I have applied many times

[shinchan79]

Description

I upgrade Terraform from v0.12 to v0.13, and encounter the following:

Terraform will perform the following actions:

  # aws_db_parameter_group.rds_parameter_group["rds_instance_001"] will be updated in-place
  ~ resource "aws_db_parameter_group" "rds_parameter_group" {
        id          = "my-web-rds-pg"
        name        = "my-web-rds-pg"
        tags        = {}
        # (4 unchanged attributes hidden)

      + parameter {
          + apply_method = "immediate"
          + name         = "time_zone"
          + value        = "asia/tokyo"
        }
      - parameter {
          - apply_method = "pending-reboot" -> null
          - name         = "time_zone" -> null
          - value        = "Asia/Tokyo" -> null
        }
        # (7 unchanged blocks hidden)
    }

Plan: 0 to add, 1 to change, 0 to destroy.

My Terraform configuration file:

resource "aws_db_parameter_group" "rds_parameter_group" {
  for_each = var.rds_instances
  name     = "${each.value.instance_name}-pg"
  family   = each.value.family

  parameter {
    name  = "character_set_client"
    value = "utf8"
  }

  parameter {
    name  = "character_set_connection"
    value = "utf8"
  }

  parameter {
    name  = "character_set_database"
    value = "utf8"
  }

  parameter {
    name  = "character_set_filesystem"
    value = "utf8"
  }

  parameter {
    name  = "character_set_results"
    value = "utf8"
  }

  parameter {
    name  = "character_set_server"
    value = "utf8"
  }

  parameter {
    name  = "log_bin_trust_function_creators"
    value = "1"
  }

  parameter {
    name  = "time_zone"
    value = "asia/tokyo"
  }
}

Expected behavior:

After I ran terraform apply, the changes will disappear, Terraform show message "No changes. Your infrastructure matches the configuration."

Actual behaviour:

After I apply, I run terraform plan again and RDS parameter group keep being modified like above.

Steps to Reproduce

1. terraform apply
2. terraform plan

[andrewpage]

Having the same problem here.

[Schnitzel]

got the same problem, time_zone is a dynamic setting in AWS:

so this should be apply_method = "immediate" but even setting this in terraform directly, somehow the state always thinks that it's apply_method = "pending-reboot" which would be a static setting and then shows a change in a plan.

This happens to us also with:

• innodb_lock_wait_timeout
• table_open_cache
• table_definition_cache
• key_buffer_size
  and many more.

[Schnitzel]

also setting apply_method = "pending-reboot" on the affected parameters removes the change in the next terraform apply, but I feel this is more a workaround as these parameters should not need pending-reboot

[matthewtckr]

I'm seeing this with v3.71.0 all the way back to v3.55.0. It does not occur in v3.54.0 or v3.46.0 (the version I was mainly using until trying to update the provider this week).

It looks like it may be caused by #18818, committed and merged starting in v3.55.0, but I'm not very familiar with the provider code.

[gclough]

Dupe of #20660 ?

[james-valente-simplisafe]

I believe this issue can be closed now.

See the merged PR #24737 for a demonstration of why this is a behavioral artifact of the AWS API and not Terraform itself. The PR adds a NOTE to the latest aws_db_parameter_group resource documentation, explaining why this is happening.

[justinretzolk]

Hey y'all 👋 Based on the details @jvalente11 included in their PR -- which added some information around this to the resource documentation -- this looks to be an upstream issue with AWS. Since the behavior is now documented, we'll close this issue for now, but I'll also be following up with a support ticket to AWS to report the unexpected behavior.

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.