Kind cluster creation ended with error: Failed to create kind cluster. No active cluster

[odockal]

Bug description

Kind cluster creation on Mac os 1.1.0 ended up with a "Cannot create kind cluster: No Active cluster". Cluster was eventually created on the filesystem. Can be verified by kind binary. But podman desktop did not detected it.

No kind cluster present:

Cli output:

Operating system

macOS 13 (Intel)

Version

1.1.0

Steps to reproduce

Create rootfull podman machine, download a kind and try to create kind cluster.

Relevant log output

No response

Additional context

I have seen the issue multiple time. I recall it was something unstable. I failed to find an issue reported, so here it is.

[jeffmaury]

Looks like a sync issue on the kubeconfig file

[odockal]

I am betting that the issue was with a "wrong" connection that was used by the podman. It was not root connection but the default one for given machine. Probably caused by #2582

[jeffmaury]

No I was able to reproduce seems to happen when you have no kubeconfig before you create the Kind cluster

[odockal]

Nice catch! Should we overcome this error (since error is there, but follow up actions will create proper kubeconfig) or keep this as know issue/feature coming from kind/kube context client?

[JurrianFahner]

No I was able to reproduce seems to happen when you have no kubeconfig before you create the Kind cluster

What @jeffmaury wrote, is exactly also the case for me. Solution would be:

1. automatic install of kubectl (in case there ~.kube\config file does not exist)
2. documentation to ask the user to install kubectl, on installation the ~.kube\config file will be automatically created.

The issue can be reproduced by removing the kind cluster and the .kube folder from the user folder.

[Blaimi]

this happens, when no ~/.kube/config-file exists when podman Desktop starts and the ingress controller shall be installed. If the file is missing, the kube-context-plugin somehow disables itself. There is also no context or "no context" lower left. The problem does not occur if at least an empty ~/.kube/context-file exists before starting podman Desktop.

The kind command works as it should: creates the cluster and sets up the context in ~/.kube/context. But kubernetes-client#createResources(…) cannot create the resources for contour because the kube-context-plugin has already disabled itself.

As a workaround execute the following commands in bash respectively PowerShell

mkdir -p ~/.kube && touch ~/.kubeconfig

If(-Not (Test-Path ~\.kube)) {New-Item -Path ~\.kube -ItemType Directory}
If(-Not (Test-Path ~\.kube\config)) {New-Item -Path ~\.kube\config -ItemType File}

[jeffmaury]

@odockal I cannot reproduce with 1.8.0 on Win11Pro. Is it MacOS specific ?

[Blaimi]

cannot reproduce this with version 1.8.0 on fedora 39 either. The cluster gets created and contour gets installed. #6406 still occurs but I think this bug can be closed as fixed.

[nichjones1]

@odockal can you retest this issue?

[odockal]

Seems like the issue was fixed. I got no error. @Blaimi seems about right, context was not updated as in mentioned issue. Still closing.