Add terraform mapping to k8s provider

[Frassle]

Proposed changes

This returns a basic mapping description for the terraform converter to map terraform programs using https://registry.terraform.io/providers/hashicorp/kubernetes to our kubernetes provider, even though it's not bridged.

I grabbed a list of all terraform kubernetes resources by running terraform providers schema -json in a tiny terraform program that just defined a kubernetes provider. I've then hooked that up to gen-kubernetes to do a best guess of mapping everything (well just resources so far) from terraform to our provider. Most things look to map pretty well by convention, but we'll probably need some manual fixups to cover everything (already got one for the "container" to containers" rename).

Related issues (optional)

[github-actions]

Does the PR have any schema changes?

Looking good! No breaking changes found.
No new resources/functions.

[Frassle]

With this mapping in place the following terraform:

provider "kubernetes" {
}

resource "kubernetes_deployment" "nginx" {
  metadata {
    name = "scalable-nginx-example"
    labels = {
      App = "ScalableNginxExample"
    }
  }

  spec {
    replicas = 2
    selector {
      match_labels = {
        App = "ScalableNginxExample"
      }
    }
    template {
      metadata {
        labels = {
          App = "ScalableNginxExample"
        }
      }
      spec {
        container {
          image = "nginx:1.7.8"
          name  = "example"

          port {
            container_port = 80
          }

          resources {
            limits = {
              cpu    = "0.5"
              memory = "512Mi"
            }
            requests = {
              cpu    = "250m"
              memory = "50Mi"
            }
          }
        }
      }
    }
  }
}

resource "kubernetes_service" "nginx" {
  metadata {
    name = "nginx-example"
  }
  spec {
    selector = {
      App = kubernetes_deployment.nginx.spec.0.template.0.metadata[0].labels.App
    }
    port {
      node_port   = 30201
      port        = 80
      target_port = 80
    }

    type = "NodePort"
  }
}

Get converted to the following TypeScript:

import * as pulumi from "@pulumi/pulumi";
import * as kubernetes from "@pulumi/kubernetes";

const nginx = new kubernetes.apps.v1beta1.Deployment("nginx", {
    metadata: {
        name: "scalable-nginx-example",
        labels: {
            app: "ScalableNginxExample",
        },
    },
    spec: {
        replicas: 2,
        selector: {
            matchLabels: {
                app: "ScalableNginxExample",
            },
        },
        template: {
            metadata: {
                labels: {
                    app: "ScalableNginxExample",
                },
            },
            spec: {
                containers: [{
                    image: "nginx:1.7.8",
                    name: "example",
                    ports: [{
                        containerPort: 80,
                    }],
                    resources: {
                        limits: {
                            cpu: "0.5",
                            memory: "512Mi",
                        },
                        requests: {
                            cpu: "250m",
                            memory: "50Mi",
                        },
                    },
                }],
            },
        },
    },
});
const nginxResource = new kubernetes.core.v1.Service("nginxResource", {
    metadata: {
        name: "nginx-example",
    },
    spec: {
        selector: {
            app: nginx.spec.apply(spec => spec?.template?.metadata?.labels?.app),
        },
        ports: [{
            nodePort: 30201,
            port: 80,
            targetPort: 80,
        }],
        type: "NodePort",
    },
});

[lblackstone]

That conversion looks good except for the selected resource version. v1beta1.Deployment is deprecated and is only retained for backward compat on very old cluster versions.

It should use v1.Deployment instead

[Frassle]

Ah cool I wasn't sure because TF had both kubernetes_deployment_v1 and kubernetes_deployment. I wasn't sure if the non versioned one should go to the first version released or the latest.

[lblackstone]

Ah cool I wasn't sure because TF had both kubernetes_deployment_v1 and kubernetes_deployment. I wasn't sure if the non versioned one should go to the first version released or the latest.

I think picking the latest version should generally be safe/backward compatible. Lets do that for now, and can revise if we run into problems.

[lblackstone]

Changes LGTM, but is there any testing for this?

[Frassle]

Changes LGTM, but is there any testing for this?

No, I wasn't sure what the best way to test this was?
We can write a .tf program and call convert on it, but should we convert to PCL and check the output of that or would y'all rather we pick a language (like typescript) and check it coverts all the way to that?

[lblackstone]

No, I wasn't sure what the best way to test this was?

cc @t0yv0 @iwahbe @rquitales for thoughts

[iwahbe]

YAML is fast, easy to target and aggressively type checked. HCL -> YAML and then previewing the yaml sounds like a great test.

[iwahbe]

Suggested change

	type TerraformAttributeSchema struct {
	}
	type TerraformAttributeSchema struct {}

[Frassle]

previewing the yaml sounds like a great test.

As in pulumi preview? Does that not need a k8s cluster spun up to run against?

[iwahbe]

previewing the yaml sounds like a great test.

As in pulumi preview? Does that not need a k8s cluster spun up to run against?

Why would pulumi preview require existing infrastructure? I'm not saying we should deploy anything, just use pulumi preview to perform a type check.

[Frassle]

Why would pulumi preview require existing infrastructure?

Same reason an aws preview complains if your not yet logged into aws.