@XinShuYang have you tried to define the KUBECONFIG env and rerun? I think this is expected when there is no right config file.

@XinShuYang can you prioritize the bug fix for this issue in Antrea 2.3?

@XinShuYang have you tried to define the KUBECONFIG env and rerun? I think this is expected when there is no right config file.

@luolanzone I believe the root cause is the expired kubeconfig file on the Windows testbed. On a new created Kubernetes cluster, I cannot reproduce this issue, but after modifying or removing the kubeconfig file, I can reproduce the same error.

The KUBECONFIG env is not mandatory. According to the current implementation, the correct kubeconfig path can be obtained from clientcmd.RecommendedHomeFile if KUBECONFIG is not set(actually it has never been set on our windows testbed). Therefore, the root cause should be the unavailability of the kubeconfig file itself.

@XinShuYang is possible to detect that antctl is out of a Pod and skip checking in-cluster config? just return kubeconfig file not found error directly if it's outside of a Pod

@XinShuYang is possible to detect that antctl is out of a Pod and skip checking in-cluster config? just return kubeconfig file not found error directly if it's outside of a Pod

@luolanzone Do you mean we will drop the support to run antctl on VM (in-cluster config is detected if running inside Pod)?

@XinShuYang is possible to detect that antctl is out of a Pod and skip checking in-cluster config? just return kubeconfig file not found error directly if it's outside of a Pod

@luolanzone In-cluster config method was added in #3645 , so even without kubeconfig file the client will still try to get in-cluster config. But I think we can enhance the error log for debugging purpose.