Kubeconfig Format Corruption After Switching Clusters and Restarting Workspace

[eye0fra]

Describe the bug

Hello,
I've encountered an issue with Kubernetes context renewal/injection. Initially, the kubeconfig is in JSON format. However, when I use the oc command line to log into a different cluster for testing, the format switches to YAML.
Upon restarting the workspace, the injection process disrupts the YAML file, causing it to become a single inline entry. Consequently, the kubeconfig becomes unusable, affecting all related dependencies, such as kubedock.
The current workaround is to delete the kubeconfig and restart the workspace to restore the correct kubeconfig format.

Che version

7.94@latest

Steps to reproduce

1. Initial Setup:

   • Ensure you have a kubeconfing in JSON format at the start of the workspaces
   • Verify that your workspace is configured to use this kubeconfig. (kubedock is working)

2. Switch Cluster:

   • Use the oc command line tool to log into a different OpenShift cluster:

     oc login <cluster-url> --token=<your-token>

   • Note that the kubeconfig format changes to YAML.
   • I believe is not necessary to use a different cluster, but run oc login command to override the config.

3. Restart Workspace:

   • Restart your development workspace.

4. Observe the Issue:

   • Check the kubeconfig file. It should now be a single inline YAML entry.
   • Attempt to use Kubernetes-related commands or dependencies like kubedock and observe failures due to the malformed kubeconfig.

5. Workaround:

   • Delete the kubeconfig.
   • Restart the workspace to restore the correct kubeconfig format.

Expected behavior

The kubeconfig should maintain its format and functionality after switching context.

Runtime

OpenShift

Screenshots

Installation method

OperatorHub

Environment

other (please specify in additional context)

Eclipse Che Logs

No response

Additional context

OpenShift DevSpaces 3.16.1

[vinokurig]

Hello @eye0fra, could you please elaborate more on:

• workspace you started, can you provide the devfile you used?
• Verify that your workspace is configured to use this kubeconfig. (kubedock is working), could you please provide some description related to kubedoc.
• Where did you launch the oc whoami command? was it your local machine terminal, or workspace pod's terminal?
  Currently I was able to reproduce the case where the config file changes from JSON to YAML in the workspace dev container terminal, but that did not affect the oc commands.

[eye0fra]

1. The workspace can be empty and is not dependent on the devfile definition.
2. If the kubedock process is running without errors, it indicates that the kube config file is intact and not corrupted.
3. What do you mean "my local machine," all the tests are on the workspace terminal. If you can execute the command afterwards, is the YAML formatted correctly? Based on the attached picture, I have doubts about its functionality.

[vinokurig]

@eye0fra According to the error this is a formatting issue, can you provide the content of the /home/.kube/config file after the new oc login. You can hide the sensitive data with ***.

[vinokurig]

@eye0fra Could you please provide the cluster's kubernetes version as well.