bug: Unable to connect to the server: getting credentials

[Keyruu]

Describe the bug
When I start kubectl.nvim it throws this error:
Unable to connect to the server: getting credentials: exec: executable aws failed with exit code 255
Error: The config profile (eks-cluster) could not be found
My guess is that my PATH isn't being respected or something.

To Reproduce

1. Use kubectl.nvim with an EKS cluster

Expected behavior
It should not throw the error

Environment (please complete the following information):

• OS: OSX
• Neovim version 0.10.1

Additional context
I am using the fish shell for shell usage, if that matters. I don't have anything special provided to the plugin just: require("kubectl").setup().

[Ramilito]

Your guess is correct!
We use session instances without any environment variables set except KUBECONFIG and PATH, mostly so we don't get any conflicts but also for safety/privacy.
You can set the aws profile env like this:

require('kubectl').setup {
  kubectl_cmd = {
    env = {
      KUBECONFIG = "$XDG_CONFIG_HOME/kube/config",
      KUBECACHEDIR = "$XDG_CACHE_HOME/kube",
    }
  }
}

[Keyruu]

I actually had some issue with my auth to AWS. But thank you for the help!

[dc-jbw]

I have a similar issue, albeit only on macOS, but it doesn't go away even if I explicitly configure the plugin environment to point to my proper K8s config and cache. The error I get is:

Unable to locate credentials. You can configure credentials by running "aws configure".  

That's more than a bit surprising because I can run any kubectl command from either that same directory or even inside a terminal in the same Neovim session. But then my AWS creds are basically defined via direnv setting environment variables found in a .envrc file because I work with multiple AWS accounts from multiple folders. Is there any way to tell the plugin to respect those environment variables? Thanks in advance.

[dc-jbw]

For others who may suffer from a similar problem, I'll describe and provide my solution. I'm essentially counting on the direnv utility to configure my AWS CLI via environment variables stored in .envrc files. Those variables are part of the environment when I launch Neovim, and they're part of any terminal shell I open in Neovim, but the K8s plugin basically discards all that information when it's running commands. So my solution, shown below in a couple code snippets from my K8s file spec, relies on a function to pull those four specific environment variables and put them into a table that can subsequently be passed to the plugin's environment options at initialization.

local function get_aws_env_vars()
    local result = {}
    local varNames = { "AWS_ACCESS_KEY_ID", "AWS_SECRET_ACCESS_KEY", "AWS_DEFAULT_REGION", "AWS_DEFAULT_OUTPUT" }
    for _, var in ipairs(varNames) do
        result[var] = os.getenv(var)
    end
    return result
end

return {
    {
        "ramilito/kubectl.nvim",
        keys = {
            { "<C-k>", function () require("kubectl").toggle() end, desc = "Toggle K8s", mode = "n" },
        },
        opts = {
            kubectl_cmd = {
                cmd = "kubectl",
                env = get_aws_env_vars(),
            }
        },
    },
}

I hope that helps somebody else. Cheers!

[Ramilito]

Awesomely solved 👏
I will write this to the wiki as well.

On another note, the choice to not start sub processes with inherited envs is out of respect for the user and a trust thing, we could offer an opt out of it through configuration if there is interest .

Maybe it's just me but I would feel safer if my programs didn't look or used my env secrets without explicitly being allowed to.