Terraform-provider-digitalocean: Changing ip address in digitalocean_domain erases all digitalocean_record s

[rca]

Terraform Version

Run terraform -v to show the version. If you are not running the latest version of Terraform, please upgrade because your issue may have already been fixed.

$ terraform -v
Terraform v0.10.8

Affected Resource(s)

• digitalocean_domain
• digitalocean_record

If this issue appears to affect multiple resources, it may be an issue with Terraform's core, so please mention this.

Terraform Configuration Files

resource "digitalocean_domain" "example" {
  name       = "example.com"
  ip_address = "1.2.3.4"
}

resource "digitalocean_record" "bar" {
  domain = "${digitalocean_domain.example.name}"

  count = "${length(digitalocean_droplet.droplets.*.id)}"

  type   = "A"
  name   = "${element(digitalocean_droplet.droplets.*.name, count.index)}"
  ttl    = 600
  value  = "${element(digitalocean_droplet.droplets.*.ipv4_address_private, count.index)}"
}

Expected Behavior

when i have the domain example.com with all its "A" records set and I change the ip_address of the digitalocean_domain resource, i expect all the digitalocean_record resources to be left or re-created if necessary.

Actual Behavior

when the digitalocean_domain's ip_address is changed, the entire resource is destroyed, which means all the digitalocean_record items are also lost. but the plan does not catch this and the DNS records are not re-created until a subsequent terraform run. This is not good, especially if you're not paying attention and DNS is left incomplete. :/

An execution plan has been generated and is shown below.
Resource actions are indicated with the following symbols:
-/+ destroy and then create replacement

Terraform will perform the following actions:

-/+ digitalocean_domain.example (new resource required)
      id:         "example.com" => <computed> (forces new resource)
      ip_address: "1.2.3.4" => "4.5.6.7" (forces new resource)
      name:       "example.com" => "example.com"


Plan: 1 to add, 0 to change, 1 to destroy.

Steps to Reproduce

Please list the steps required to reproduce the issue, for example:

1. terraform apply
2. change ip_address
3. terraform plan
4. terraform apply
5. see missing record entries
6. terraform plan
7. terrraform apply
8. see all the records come back

Important Factoids

Nothing special

[wrkngu0]

can confirm this behavior

[andrewsomething]

We've just released a change to the DO API that should facilitate fixing this. ip_address is now optional. Once Terraform is updated to reflect that, you can create the A record explicitly using digitalocean_record resource rather that automagically through the digitalocean_domain. See: https://developers.digitalocean.com/documentation/changelog/api-v2/create-domains-without-providing-an-ip-address/

[andrewsomething]

I'm happy to say that yesterday, we release version 1.0.0 of the DigitalOcean Terraform provider. It should include a fix for this issue.

You can find the full changelog here: https://github.com/terraform-providers/terraform-provider-digitalocean/blob/master/CHANGELOG.md#100-september-27-2018

In order to upgrade an existing project to the latest version, run:

terraform init -upgrade

If you believe that this issue is being closed by mistake or the problem has not been resolved, please let me know or feel free to open a new issue with additional details.

Thanks for your patience!