Rancher Kubernetes cluster doesn't detect the node failure

When I shutdown host, in UI it becomes in state Reconnecting, but kubectl shows
NAME STATUS AGE
rancherkub1 Ready 1d
rancherkub2 Ready 1d

when rancherkub2 is down. As a result kubernetes doesn’t reschedule containers that were running on this host.

Also I tried use k8s health checks, but it doesn’t change anything. Raw k8s cluster works normally detecting this failures.

I commented on your GH issue:

https://github.com/rancher/rancher/issues/5515