Deprecation warning for workload_metadata_config parameter

[wyardley]

I'm getting the following as permadrift with https://github.com/terraform-google-modules/terraform-google-kubernetes-engine/tree/master/modules/beta-private-cluster and Google provider v3.88.0:

~ resource "google_container_node_pool" "pools" {
[…]
   ~ node_config {
[…]
     ~ workload_metadata_config {
       - mode     = "GKE_METADATA" -> null

If I explicitly set node_metadata = "GKE_METADATA_SERVER", I get a deprecation warning about node_metadata being deprecated for mode, presumably from:

terraform-google-kubernetes-engine/modules/beta-private-cluster/cluster.tf

Line 234 in 660ddc9

node_metadata = workload_metadata_config.value.node_metadata

Either way, I can “apply” the drift, but it shows up on the next plan / apply. Is there a way to avoid this?

[morgante]

Permadiffs should be addressed at the provider. Please open a bug here: https://github.com/hashicorp/terraform-provider-google/issues

[wyardley]

@morgante won't the deprecation warning

│ Warning: Deprecated Attribute
│ 
│   with module.gke.google_container_cluster.primary,
│   on .terraform/modules/gke/modules/beta-private-cluster/cluster.tf line 227, in resource "google_container_cluster" "primary":
│  227:     node_config {
│ 
│ Deprecated in favor of mode.

be an issue in either case (esp. if the module is supporting both old and new versions of the provider)?

[wyardley]

Also wondering if I may be missing a setting needed for this to not register?

Oddly, a plan with the same provider version didn't seem to exhibit this issue a couple of hours ago when I updated to the latest provider versions 🤔

I did try pinning a version or two back of this module (as well as the provider), and neither seemed to fix it.

[morgante]

Oh sorry, I missed the deprecation warning. Yes we should fix that.

Oddly, a plan with the same provider version didn't seem to exhibit this issue a couple of hours ago when I updated to the latest provider versions 🤔

My guess is the API behavior changed.

[wyardley]

My guess is the API behavior changed.

Oh, derp. yeah, I've seen that happen before
Filed hashicorp/terraform-provider-google#10300
with some debug output that may help them figure it out

[alexku7]

I forked the module localy and changed the deprecated node_metadata to node and to the value GKE_METADATA instead of the previous GKE_METADATA_SERVER and seems the error has gone

[onimsha]

For anyone is interested, i had to pin the exact provider version in my terraform configuration like this to not having the plan drift

terraform {
  required_providers {
    google = {
      source  = "hashicorp/google"
      version = "3.78"
    }
    google-beta = {
      source  = "hashicorp/google-beta"
      version = "3.82"
    }
    kubernetes = {
      source  = "hashicorp/kubernetes"
      version = "1.13.4"
    }
  }
}

[brandondtb]

FYI, regarding version pinning, I only needed to pin the google provider to the version prior to the latest, 3.87.0 in order to avoid these updates.

[wyardley]

Hrm. I tried pinning one version back and didn’t have any luck, but will try it again.

[FearlessHyena]

3.87.0 worked for me too
Also seems like hashicorp/terraform-provider-google#10300 submitted by @wyardley has been fixed so hopefully the next provider release will include the fix

[JCPistell]

FYI, upgrading to 3.90.0 does indeed seem to resolve this issue, but now I'm getting a deprecation warning for basic auth. Reported here: #1042

[JCPistell]

The issue has been fixed in 3.90.1: hashicorp/terraform-provider-google#10479

[mike-serchenia]

using 3.90.1
gets another error: terraform-google-kubernetes-engine/issues/1045

[FalconerTC]

This is broken again in 4.0+