fix: add UntaggedNetwork and remove NetworkTypeCustom

[FrankYang0529]

issue: harvester/harvester#5226

Test

Setup: Create a Harvester cluster

Case 1: new terraform provider

1. Build terraform-harvester-provider.

CGO_ENABLED=0 go build -mod=vendor -o bin/terraform-provider-harvester .
mkdir -p ~/.terraform.d/plugins/registry.terraform.io/harvester/harvester/0.0.0-dev/linux_amd64
cp bin/terraform-provider-harvester ~/.terraform.d/plugins/registry.terraform.io/harvester/harvester/0.0.0-dev/linux_amd64/terraform-provider-harvester_v0.0.0-dev

2. Create a temporary folder and put the following content in versions.tf.

terraform {
  required_providers {
    harvester = {
      source  = "harvester/harvester"
      /* version = "0.6.4" */
      version = "0.0.0-dev"
    }
  }
}

provider "harvester" {
  kubeconfig = "~/.kube/config"
}

3. Put the following content in main.tf in the temporary folder.

data "harvester_clusternetwork" "mgmt" {
  name = "mgmt"
}

resource "harvester_network" "mgmt-untagged-network" {
  name      = "mgmt-untagged-network"
  namespace = "harvester-public"

  vlan_id = 0

  route_mode           = "auto"
  route_dhcp_server_ip = ""

  cluster_network_name = data.harvester_clusternetwork.mgmt.name
}

resource "harvester_network" "mgmt-vlan1" {
  name      = "mgmt-vlan1"
  namespace = "harvester-public"

  vlan_id = 1

  route_mode           = "auto"
  route_dhcp_server_ip = ""

  cluster_network_name = data.harvester_clusternetwork.mgmt.name
}

4. Run terraform init and terraform apply. There are two VM Networks without error.
5. Run terraform destroy to remove all data.

Case 2: Upgrade path

1. Create a temporary folder and put the following content in versions.tf.

terraform {
  required_providers {
    harvester = {
      source  = "harvester/harvester"
      version = "0.6.4"
      /* version = "0.0.0-dev" */
    }
  }
}

provider "harvester" {
  kubeconfig = "~/.kube/config"
}

2. Put the following content in main.tf in the temporary folder.

data "harvester_clusternetwork" "mgmt" {
  name = "mgmt"
}

resource "harvester_network" "mgmt-vlan1" {
  name      = "mgmt-vlan1"
  namespace = "harvester-public"

  vlan_id = 1

  route_mode           = "auto"
  route_dhcp_server_ip = ""

  cluster_network_name = data.harvester_clusternetwork.mgmt.name
}

3. Run terraform init and terraform apply. There is one VM Networks without error.
4. Build terraform-harvester-provider.

CGO_ENABLED=0 go build -mod=vendor -o bin/terraform-provider-harvester .
mkdir -p ~/.terraform.d/plugins/registry.terraform.io/harvester/harvester/0.0.0-dev/linux_amd64
cp bin/terraform-provider-harvester ~/.terraform.d/plugins/registry.terraform.io/harvester/harvester/0.0.0-dev/linux_amd64/terraform-provider-harvester_v0.0.0-dev

5. Update versions.tf to use 0.0.0-dev.
6. Run terraform init --upgrade and terraform apply. No change happen.

data.harvester_clusternetwork.mgmt: Reading...
data.harvester_clusternetwork.mgmt: Read complete after 0s [id=mgmt]
harvester_network.mgmt-vlan1: Refreshing state... [id=harvester-public/mgmt-vlan1]

No changes. Your infrastructure matches the configuration.

Terraform has compared your real infrastructure against your configuration and found no differences, so no changes are needed.

Apply complete! Resources: 0 added, 0 changed, 0 destroyed.

[starbops]

I'm not sure if I understand it correctly. The labels here seem to be a hackish way of passing values among different custom processors. And they will remain on the resulting NAD object. The mutator is responsible for populating those labels using the .spec.config field. I'd suggest we leverage the NetConf struct in the network controller project to process the NAD object better. We can then serialize it in the Result() function. But if I get the wrong perception about how the Terraform provider works, please ignore my 2c. Thank you.

[starbops]

LGTM, thank you!

[TomTucka]

@starbops Any idea when this could be merged?

[bk201]

LGTM, thanks!