google_sql_database_instance: Optional+Computed settings attributes show persistent phantom "delete" diffs when set to null

[soepic1]

Community Note

Several nested Optional+Computed attributes inside settings — e.g.
collation, time_zone, backup_configuration.location,
final_backup_config.retention_days, location_preference.follow_gae_application,
location_preference.secondary_zone, ip_configuration.allocated_ip_range,
ip_configuration.server_ca_pool, sql_server_audit_config.bucket,
replication_cluster.failover_dr_replica_name, and the
read_pool_auto_scale_config numeric fields — return empty string "" / 0
from the Cloud SQL Admin API when not explicitly configured.

When these are set to null (the natural way to express "don't manage
this") in Terraform config, terraform plan shows a delete action for
these attributes on every single plan, even though the underlying GCP
value never changes. This creates a permanently "dirty" plan.

Terraform Version & Provider Version(s)

Terraform Version

Terraform v1.11.4
+ provider registry.terraform.io/hashicorp/google v5.44.0

### Affected Resource(s)

google_sql_database_instance

### Terraform Configuration

terraform {
required_providers {
google = {
source = "hashicorp/google"
version = "~> 5.0"
}
}
}

provider "google" {
project = "your-gcp-project-id"
region = "europe-west2"
}

resource "google_sql_database_instance" "repro_instance" {
name = "repro-sql-instance-null-diff"
database_version = "MYSQL_8_0_31"
region = "europe-west2"

settings {
tier = "db-perf-optimized-N-8"
collation = null
time_zone = null

backup_configuration {
  enabled  = true
  location = null
}

ip_configuration {
  allocated_ip_range = null
  server_ca_pool     = null
}

}
}

### Debug Output

_No response_

### Expected Behavior

Since these are Optional+Computed, no diff should be shown when config
leaves them unset and state already matches the provider's zero-value
default.

### Actual Behavior

```text
  # google_sql_database_instance.repro_instance will be updated in-place
  ~ resource "google_sql_database_instance" "repro_instance" {
        id   = "repro-sql-instance-null-diff"
        name = "repro-sql-instance-null-diff"

      ~ settings {
          - collation = "" -> null
          - time_zone = "" -> null

          ~ backup_configuration {
              - location = "" -> null
            # (2 unchanged attributes hidden)
          }

          ~ ip_configuration {
              - allocated_ip_range = "" -> null
              - server_ca_pool     = "" -> null
            # (4 unchanged attributes hidden)
          }
        }
    }

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

Steps to reproduce

1. Save the minimal configuration above into main.tf.
2. Run terraform init and terraform apply to provision the instance.
3. Immediately run a subsequent terraform plan without making any changes to main.tf or the GCP console.
4. Observe that terraform plan outputs a non-zero plan proposing to delete collation, time_zone, location, allocated_ip_range, and server_ca_pool from "" to null.

Important Factoids

This behavior is particularly impactful on production databases managed via CI/CD automation pipelines (such as Terraform Cloud), where permanent non-zero diffs make it difficult to verify zero-downtime infrastructure reconciliations.

References

No response

b/558324049

[ggtisc]

Confirmed permadiff issue with the following code:

resource "google_sql_database_instance" "sql_db_instance_29237" {
  name                = "sql-db-instance-29237"
  database_version    = "MYSQL_8_0_31"
  region              = "europe-west2"
  deletion_protection = false

  settings {
    tier      = "db-perf-optimized-N-8"
    edition   = "ENTERPRISE_PLUS"
    collation = null
    time_zone = null

    backup_configuration {
      enabled  = true
      location = null
    }
  }
}