Provider using local config despite paramaters

[andrew-gallin-current]

My provider appears to have started exclusively using the local config regardless of provider configuration. This only just started and I have been using this for several months.

The active cluster is used if not specified, but if I explicitly set the config profile param, the provider will use that.

I have removed the installed provider, reinstalled, removed and reinitialized my .terraform file and restarted my machine and the issue still persists.

This is the way my provider is invoked

terraform {
  required_version = ">= 1"
  required_providers {
    google = {
      source  = "hashicorp/google"
      version = ">= 4.27.0, < 5.0.0"
    }
    kubectl = {
      source  = "gavinbunney/kubectl"
      version = ">= 1.14.0"
    }
  }
}

data "google_client_config" "default" {
  provider = google.tokengen
}

data "google_container_cluster" "CLUSTER" {
  name     = var.cluster_name
  location = var.region
}

provider "kubectl" {
  load_config_file = false
  host             = "https://${data.google_container_cluster.CLUSTER.endpoint}"
  token            = data.google_client_config.default.access_token
  cluster_ca_certificate = base64decode(
    data.google_container_cluster.CLUSTER.master_auth[0].cluster_ca_certificate,
  )
}

Here are some debug logs:

2022-11-21T15:31:43.425-0500 [DEBUG] provider.terraform-provider-random_v3.4.3_x5: Called provider defined Provider GetDataSources: tf_provider_addr=registry.terraform.io/hashicorp/random tf_req_id=0d6f8408-ebdd-997e-0ebc-8e27d6c67940 tf_rpc=GetProviderSchema @caller=github.com/hashicorp/terraform-plugin-framework@v0.11.1/internal/fwserver/server.go:195 @module=sdk.framework timestamp=2022-11-21T15:31:43.425-0500
2022-11-21T15:31:43.425-0500 [DEBUG] provider.terraform-provider-kubectl_v1.14.0: 
2022/11/21 15:31:43 [DEBUG] Using kubeconfig: /Users/andrewgallin/.kube/config
2022-11-21T15:31:43.425-0500 [DEBUG] provider.terraform-provider-kubectl_v1.14.0: 
2022/11/21 15:31:43 [WARN] Invalid provider configuration was supplied. Provider operations likely to fail: stat /Users/andrewgallin/.kube/config: no such file or directory

Then I auth to the cluster with a  gcloud container clusters get-credentials CLUSTER --zone=us-central1

2022-11-21T15:35:36.123-0500 [DEBUG] provider.terraform-provider-kubectl_v1.14.0: 
2022/11/21 15:35:36 [DEBUG] Using kubeconfig: /Users/andrewgallin/.kube/config

Any idea what may be going on? The provider.go file seems to indicate that as long as it is set to false it would skip the config path.

I think using the local config by default should be turned off and people should opt in regardless. Not sure that would help this bug as it is ignoring that "false" param, but maybe it would.

[MalteKud]

I had a similar issue and found that the last working version seems to be 1.10.0.

[jamesosbourn]

Having similar issues trying to configure MetalLB resources using this provider. I was eventually able to get this to work after reading this and downgrading from the latest version to 1.11.0. I tried all of the 1.11.x version in reverse order and this was the first version that worked. I had also tried 1.10.0 as per the comment above which worked.

Getting the same error that the local config file does not exist and a failure to talk to the k8s cluster.