feat(server): allow rebuilding with image and/or user-data

[svanrossem]

Due to current capacity constraints in several locations, this change adds support for rebuilding an existing node with a new image, ISO, or updated user-data. This provides a practical alternative when provisioning a new node in the desired location is not possible.

With the hcloud-go dependency updated to v2.36.0 (including the changes from hcloud-go PR #813), rebuild functionality is now available and can be leveraged by the provider.

I'm also thinking about a rebuild-argument for rebuilding an existing node without updating the image/ISO or user-data?

[venkatamutyala]

A rebuild argument would be very helpful. This would be very useful for situations where there isn't an image change but people want to start over. Some usecases would be to re-apply an immutable setup or when an ansible playbook requires a clean deployment

[apricote]

I am hesitant to accept the pull request in the current form. Please take a look at my concerns in #1348 (comment)

[svanrossem]

resource "cloudinit_config" "foobar" {
  gzip          = false
  base64_encode = false

  part {
    filename     = "hello-script.sh"
    content_type = "text/x-shellscript"

    content = file("${path.module}/hello-script.sh")
  }

  lifecycle {
    action_trigger {
      events  = [after_create]
      actions = [action.hcloud_server_rebuild.rebuild]
    }
  }
}

resource "hcloud_server" "foo" {
  name        = "foo"
  image       = "fedora-43"
  server_type = "cx23"
  location    = "fsn1"
  user_data   = cloudinit_config.foobar.rendered
  rebuild_when_necessary = true

  public_net {
    ipv4_enabled = false
    ipv6_enabled = true
  }

  lifecycle {
    ignore_changes = [user_data]
  }
}

/* $ terraform apply -invoke action.hcloud_server_rebuild.rebuild */
action "hcloud_server_rebuild" "rebuild" {
  config {
    server_id = hcloud_server.foo.id
    
    image     = hcloud_server.foo.image
    user_data = hcloud_server.foo.user_data
  }
}