Switch to version 1.0.0 : node-fetch behind proxy

[Leer0r]

Describe the bug
I'm currently using the 1.0.0 version, and I notice on the changelog that the project switched from request to node-fetch for APIs call. With request I can use HTTP_PROXY and HTTPS_PROXY variables to access to my cluster, but the new package doesn't allowed to do such things. In my research, I've seen a hint of @lselden who is using the global-agent package to import the proxy setting. I use it and my client can now reach the kubernetes API, but I have the Unauthorized error, so I suppose that some kubeconfig authorization are not in the forwarded request.

Client Version
1.0.0

Server Version
1.31.3

To Reproduce

1. Install @kubernetes/client-node stable version on a nodeJS project behind a proxy
2. Install global-agent package and bootstrap it
3. Create a kubernetes API with makeApiClient
4. Try to use any fetch method from the client API

Expected behavior
The API return the expected information by reaching the cluster

Example Code
Simple code

import {bootstrap} from 'global-agent'
import k8s, {CoreV1Api, Metrics, V1NodeCondition} from '@kubernetes/client-node'

//PROXY settings
process.env.NODE_TLS_REJECT_UNAUTHORIZED = "0";
process.env["GLOBAL_AGENT_HTTP_PROXY"] = process.env.HTTP_PROXY || "http://my.proxy.address:my.proxy.port";
bootstrap({
  forceGlobalAgent: true
})

const config = new k8s.KubeConfig()
config.loadFromDefault()
const kube_api = config.makeApiClient(k8s.CoreV1Api)

kube_api.listNode()

Environment (please complete the following information):

• OS: Ubuntu (WSL) and Ubuntu (bare metal)
• Node.js version v23.3.0
• Cloud runtime On-prem K3S install

Additional context
When I use the same kubeconfig file with kubectl, I can access to the cluster

Here I'm on my local WSL, accessing the remote cluster

[mstruebing]

Does it work if you set the proxy configuration in your KUBECONFIG?

Reference:

• Add proxy support #2111
• proxy-url in kube config not used #853

[Leer0r]

Adding the proxy-url field in the kubeconfig file seems not handle by the client, I have the same issue than before. Regarding the small hack from @geisbruch, it not work due to a missing property:

Here is the plain code if you need it

type ApiConstructor<T extends ApiType> = new (config: Configuration) => T;
function connectToKubeClusterWithProxy(proxyURL: string) {
    const kc = new k8s.KubeConfig();
    const originalMakeApiClient = kc.makeApiClient;
    kc.makeApiClient = function <T extends ApiType>(api: ApiConstructor<T>): T {
        const client = originalMakeApiClient.call(kc, api);
        client.interceptors.push(async (config: Configuration) => {
            config.proxy=proxyURL;
        });
        return client as T;
    };
}

Also the ApiConstructor interface is not exported by default, so I need to recreate it in my code

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[k8s-triage-robot]

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle rotten
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle rotten