Endless Reconcile Loop in RDS Controller Due to Subnet Order Difference in DBSubnetGroup CRD

[doryer]

Describe the bug
Recently, we performed a migration from Terraform-provisioned RDS to RDS Operator using AdoptedResource CRDs. We had multiple DBs with different subnet groups, and the order of the three subnets we were using (across three AZs) was different. We wanted to align them to match the AZ order. Instead of having:

• Subnet-AZ2

• Subnet-AZ1

• Subnet-AZ3
  We aimed to have:

• Subnet-AZ1

• Subnet-AZ2

• Subnet-AZ3
  When we created the adopted resource and then synced the DBSubnetGroup new CRD, the difference was only in the list order, but the subnets were the same. I expected no changes to occur. However, the RDS controller logs indicated that it treated these as six different subnets (three from the first list and three from the second list). Logs:

{ "level": "info", "ts": "2024-06-04T12:47:06.520Z", "logger": "ackrt", "msg": "desired resource state has changed", "kind": "DBSubnetGroup", "namespace": "default", "name": "dummy-sgn", "account": "123456789", "role": "", "region": "eu-west-1", "is_adopted": false, "generation": 1, "diff": [ { "Path": { "Parts": [ "Spec", "SubnetIDs" ] }, "A": [ "subnet-B", "subnet-A", "subnet-C" ], "B": [ "subnet-B", "subnet-A", "subnet-C", "subnet-C", "subnet-B", "subnet-A" ] } ] }

Steps to reproduce

1. Create a DBSubnetGroup CRD using AdoptedResource based on an existing SubnetGroup.
2. Apply the same list of subnets, but in a different order.

Expected outcome
The RDS Controller should recognize that these are the same three subnets, so no changes should be performed, avoiding an endless reconcile loop.

Environment

• Kubernetes version - 1.25
• Using EKS (yes/no), if so version? Yes, EKS 1.25
• AWS service targeted (S3, RDS, etc.) - RDS