Update to support kubernetes provider v2

[woz5999]

Breaking Changes

• This module requires kubernetes Terraform provider >= 2.0, which introduced breaking changes to the provider. This module has made additional breaking changes in response.
• Inputkubernetes_load_config_file has been removed because the corresponding load_config_file option to the Kubernetes provider has been removed in provider version 2.0.
• Input kubernetes_config_path has been removed as its use case (Allow k8s config file load #77, feat: add k8s provider config variables #78) was for one-time migration and in general we want to discourage its use. If needed anyway, a value can be provided via the KUBE_CONFIG_PATHS environment variable. If a more compelling use case is made, it will be a non-breaking change to restore it.
• Users of previous versions of this module can continue to use those versions by pinning the Kubernetes provider version in their root module. For Terraform >= 0.12.26, add this block to your required_providers block

    kubernetes = {
      source  = "hashicorp/kubernetes"
      version = "~> 1.11"
    }

what

• Update to support kubernetes provider v2

why

• Better. Stronger. Faster.

references

• https://github.com/hashicorp/terraform-provider-kubernetes/releases/tag/v2.0.0
• closes chore(deps): update terraform kubernetes to v2 #95

[woz5999]

Due to provider behavior changes, this value no longer has a default. Previously, if load_config_file, was false, this config was ignored. Updating the default to match the provider's maintains the same behavior as before.

[woz5999]

/test all

[vsimon]

Just hit this a couple minutes ago 😄 thanks for opening a PR

[Nuru]

I am not convinced this is really backwards compatible. The new Kubernetes provider drops support for KUBECONFIG, load_config_file, and the default ~/.kube/config as a default config_path. What will this break, how will clients fix it?

[woz5999]

@Nuru load_config_file was always kind of redundant with config_file. The new provider behavior does the logical thing and effectively determines whether or not to load the config file based on whether or not config_file is set. So previously, ~/.kube/config was ignored if load_config_file was false. This maintains that behavior.

For those that are loading a config file, chances are good they're already providing a path, and if they're not and relying on the default, the v2 provider would be a breaking change for them anyway. So for that segment of users, they'll have a v2 breaking change and just need to explicitly provide their path.

Leaving the default path in the variable will break this for everyone who previously had load_config_file set to false, which was the old variable default.

[mergify]

This pull request is now in conflict. Could you fix it @woz5999? 🙏

This Pull Request has been updated, so we're dismissing all reviews.

[woz5999]

/test all

[Nuru]

I think we can make this backwards compatible instead. Hang on please while I investigate.

This Pull Request has been updated, so we're dismissing all reviews.

[Nuru]

/test all

@woz5999 after consulting with the team, we decide just to remove these variables altogether as they have no good use case for this module, and then move forward with the Kubernetes provider version pin >= 2.0. Thank you for your efforts and for your patience.