Tags for digital ocean droplets keep changing order about terraform-provider-digitalocean HOT 8 CLOSED

I'm trying to use sort() with a list of tags but it has o effect:

locals = {
  tags = ["${var.stage}", "${var.group}", "${var.env}"]
  tags_sorted = "${sort(distinct(local.tags))}"
  tags_count = "${length(local.tags_sorted)}"
}

resource "digitalocean_tag" "host" {
  name  = "${element(local.tags_sorted, count.index)}"
  count = "${local.tags_count}"
}

resource "digitalocean_droplet" "host" {
...
    tags = ["${sort(digitalocean_tag.host.*.id)}"]
...
}

The result is still the same, tags re-ordering on every run.

from terraform-provider-digitalocean.

Same issue here.

-                            "tags.0": "worker",
-                            "tags.1": "cluster",
+                            "tags.0": "cluster",
+                            "tags.1": "worker",

from terraform-provider-digitalocean.

Same issue.
Use sort() is a workaround

from terraform-provider-digitalocean.

This still seems to be an issue, @jackivanov can you elaborate and give an example?

from terraform-provider-digitalocean.

this worked for me

module "do_swarm_worker" {
source = "./do_swarm_worker"
token = "${var.do_token}"
region = "${var.do_region}"
domain = "${var.domain}"
tags = ["${sort(list(digitalocean_tag.cluster.id,digitalocean_tag.worker.id))}"]
}

from terraform-provider-digitalocean.

Very annoying issue

from terraform-provider-digitalocean.

Hello fellow DigitalOcean Terraformers! I've joined the group of people suffering from the issue. No sort work-arounds are working for me. I've recently also hit a hard limit of DigitalOcean firewalls (100), what caused me to start heavily rely on tags. So now for each TF workspace, we do get a plan with changes, even tho there are not real changes :(

Considering that we are kind of forced to use tags if we want to utilise digitalocean_firewall effectively, I think this issue is more than annoying. Any known work-arounds are welcome!

Side question - did anyone try to sort the problem with lifecycle & ignore_changes? I haven't explored it yet as it feels like dirty hack - I still kind of care about changes to tags (if there are actual changes)

from terraform-provider-digitalocean.

@paranoidd It totally depends on your implementation/requirements/needs. For example, you can do it like in this demo https://github.com/baffolobill/terraform-digitalocean-tags-hack (this is part of my project, not sure that all dependencies included) - I just manually process tags via bash scripts. This is just an idea, not complete solution.

resource "null_resource" "assign_tags_to_droplet" {
    depends_on = [
        "digitalocean_droplet.bastion",
        "module.droplet_tags",
    ]
    count      = "${var.num_nodes}"

    triggers {
        tag_ids    = "${join(",", module.droplet_tags.tags)}",
        droplet_id = "${element(digitalocean_droplet.bastion.*.id, count.index)}",
    }

    # !!! All magic does this script. You can test it like so:
    provisioner "local-exec" {
        command = "\"${path.module}/../../../../scripts/assign_tags_to_droplet.sh\" ${element(digitalocean_droplet.bastion.*.id, count.index)} ${join(" ", module.droplet_tags.tags)}"
    }
}

I do so because of another problem with tags: in some cases when Terraform sees changes in tags, it destroys a tag by deleting it (contrary to "untag") (if it's not obvious, Terraform doesn't re-add resources to a tag, when it re-creates a tag). As result, this tag disappear from all resources, even from those which aren't managed by Terraform. It brings me to situation when Droplets didn't see each other (my firewall rules totally rely on tags), consul couldn't connect to leader (because auto-join rely on tags).

from terraform-provider-digitalocean.