Terraform import (complex) - resource names

Question

I'm writing a custom Terraform provider and I'd like to implement the so-called complex case. I.e. my resource returns multiple sub-resources. When I look at the resulting .tfstate file, the names of the sub-resources are generated from the names of the parent resource. So, for example, if I have a parent resource named 'cluster', then the sub-resources are named 'cluster', 'cluster-1', 'cluster-2' etc which is pretty useless to me. I can't find any way of controlling this outcome from within the provider. Is it possible to return a sensible name per sub-resource, based on the sub-resource attributes?

Here's a pruned .tfstate example of what I mean. Note how the name fields of each my_sub_resource are derived from the my_parent_resource resource name field.

{
  "version": 4,
  "terraform_version": "1.0.1",
  "serial": 1,
  "lineage": "9e9e660b-35fb-df03-119a-5e8d06e775f1",
  "outputs": {},
  "resources": [
    {
      "mode": "managed",
      "type": "my_sub_resource",
      "name": "parent",
      "provider": "***",
      "instances": [
        {
          "schema_version": 0,
          "attributes": {
            "parent": "e2e7c3c2-219d-4ab1-a981-330cc431c0df",
            "id": "9e2799fc-44d7-45eb-95bc-a8dd34c45ba1",
            "tag": "resource1"
          },
          "sensitive_attributes": []
        }
      ]
    },
    {
      "mode": "managed",
      "type": "my_sub_resource",
      "name": "parent-1",
      "provider": "***",
      "instances": [
        {
          "schema_version": 0,
          "attributes": {
            "parent": "e2e7c3c2-219d-4ab1-a981-330cc431c0df",
            "id": "2f4fa368-b7b2-44cb-a23d-ddb89696d55f",
            "tag": "resource2"
          },
          "sensitive_attributes": []
        }
      ]
    },
    {
      "mode": "managed",
      "type": "my_parent_resource",
      "name": "parent",
      "provider": "***",
      "instances": [
        {
          "schema_version": 0,
          "attributes": {
            "id": "e2e7c3c2-219d-4ab1-a981-330cc431c0df",
            "name": "parent"
          },
          "sensitive_attributes": []
        }
      ]
    }
  ]
}

Answer 1

To my knowledge this can't be fixed. What I did as a workaround, instead of having 2 separate resources, was have a single resource with multiple sub resource blocks. Each of the sub-resources can have a name attribute and I therefore have the level of control I need. E.g.

resource "provider_cluster" "cluster" {
  cluster_app {
    name = "app1"
  }
  cluster_app {
    name = "app2"
  }
  cluster_app {
    name = "app3"
  }
}