kind + porter + kubeconfig: connect: connection refused

[snocke]

Hi there,
I'm using an M1 and I'm trying to use porter with my KinD Cluster. I've followed the https://porter.sh/best-practices/kind/ instructions. However, porter is always trying to connect to the default IP and not the actual ~/.kube/config
kubectl cluster-info
Kubernetes control plane is running at https://10.15.19.162:6443
CoreDNS is running at https://10.15.19.162:6443/api/v1/namespaces/kube-system/services/kube-dns:dns/proxy

Executing porter install -c kubeconfig results in:
unable to recognize "kubernetes/manifests/hbase.yml": Get http://localhost:8080/api?timeout=32s: dial tcp 127.0.0.1:8080: connect: connection refused

I've also tried the new kubeConfig (https://github.com/getporter/kubernetes-mixin/blob/main/pkg/kubernetes/testdata/install-input-with-kubeconfig.yaml). However, I receive the error error: stat /Users/me/.kube/config: no such file or directory when executing porter install -c kubeconfig

I wonder about the origin of the problem. Is it due to the M1? Is it Porter? Or did I configure something wrong?
Anyone running in similar problems or got an idea for a fix?
I appreciate any help. Thanks!

[carolynvs]

Have you tried the steps from our guide to using KinD with Porter yet? https://getporter.org/best-practices/kind/

[snocke]

thanks. Yes, i tried that. Specified a kind-config.yaml with the ip adress extracted from ifconfig en0 | awk '/inet / {print $2; }' | cut -d ' ' -f 2. I'm gonna produce some logs for more insights and post it here.

[carolynvs]

This error message is usually what you see when there isn't a kubeconfig in the location where kubectl expects it, so it's attempting the default address.

unable to recognize "kubernetes/manifests/hbase.yml": Get http://localhost:8080/api?timeout=32s: dial tcp 127.0.0.1:8080: connect: connection refused

Here's what your porter.yaml should look like (below). Also make sure that you are using v0.28.5 of the kubernetes mixin:

mixins:
  - kubernetes

credentials:
  - name: kubeconfig
    # Note we changed from /home/root to /home/nonroot in a recent version of v1
    path: /home/nonroot/.kube/config

install:
  - exec:
      description: "Check connectivity"
      command: kubectl
      arguments:
        - get
        - nodes

[carolynvs]

Sorry the docs didn't explain this recent change. I've created an issue to make sure we get the docs to match

getporter/kubernetes-mixin#41

[snocke]

Hi Carolyn, thank you.
What got me confused in the docs was the sentence The paths and environment variable names used in the credential declaration represent where the value of the injected credentials are stored in the bundle when it is executing. They are not used to locate the credential, that is the responsibility of credential sets. https://porter.sh/credentials/

As far as I understand the credentials are indeed used to locate e.g. the kubeconfig.

Problem solved. Thanks!

[carolynvs]

What I meant by that sentence in the docs was that when a bundle author is specifying which credentials the bundle needs, they only get to say "I need a kubeconfig and it should be placed in /home/nonroot/.kube/config". The path specified by the bundle author in a credential is not the path on the host computer, but instead a path inside the bundle where Porter will put the credential provided by the end-user when they run the bundle.

The end-user defines a credential set which says "when the bundle needs a kubeconfig, this is where it should come from" and can use a kubeconfig from their laptop, or from the output of a command, or from a value stored in a secret vault, etc.

I'm not sure if that helps clarify what the docs for credentials meant?