A bug was recently discovered in v2.4.0 regarding node agents not reconciling correctly, which causes kubelets to constantly restart. [#26339]
Note: The workloads on your clusters will still be functional, but kubelet will constantly be restarted.
You can correct your clusters facing this issue by redeploying the cattle-node-agent
pod on your nodes to retrigger the unmanaged agent cleanup using kubectl
.
kubectl -n cattle-system delete pod -l app=cattle-agent
Rancher v2.4.1 should be released in the next day or two to address this issue.