Node loses label

[alimoli]

We are using GCP and I tried to test the configuration of kube IP by restarting a node and see if the static IP was correctly restored. The kube IP daemon does not start a new pod (that one that was previously deleted with the node stop) for the restarted node.

The issue is that, once restarted, a node loses its custom label (kubeip.com/public=true).

It sounds like a bug, but I could not find any solution to that.
I would be curious to know how others implemented it.

[ninadbstack]

We are facing same issue

[prklm10]

In V1, It is due to copylabels and clearlabels default value is true ref, which is causing the label to be missed. Make them false and this should work.