Issues running with Alpine/Docker

[nbhargava]

Describe the bug
In a docker container when initializing the k8s provider with values from a GKE instance generated by terraform, the provider will attempt to connect to a kubernetes cluster at localhost.

Steps to reproduce the issue:
See https://github.com/nbhargava/terraform-k8s-alpine. After plugging in a real GCP project name in main.tf and storing credentials in ./creds:

• Running terraform plan succeeds normally (at least on macOS)
• Running docker build . fails

The error given is:

Error: Failed to configure: Get http://localhost/api?timeout=32s: dial tcp 127.0.0.1:80: connect: connection refused

  on main.tf line 45, in provider "k8s":
  45: provider "k8s" {

Expected behavior
docker build . should succeed and not try to connect to a cluster before the host/token/certificate are known. At the very least, the behavior should be consistent across all platforms. I also tried changing from an Alpine based image to Debian Stretch, but the docker-based approach still failed.

[sagikazarmark]

Do you have the same versions of the provider locally and in docker?

Can you please try the latest stable version: https://github.com/banzaicloud/terraform-provider-k8s/releases/tag/v0.7.2

Recently, we've rewritten the provider to connect clusters the same way as the official kubernetes provider does.

[nbhargava]

Bumped terraform in both places to 0.12.21 just to be sure.

The problem is the same using the pre-built binaries (again just checking macOS vs alpine docker). Updated code at the previously linked repo.

[DesmondH0]

I have the same issue with the plugins as below:

Terraform v0.12.23
+ provider.google v3.13.0
+ provider.helm v1.1.0
+ provider.k8s v0.7.4
+ provider.kubernetes v1.11.1