Symptoms¶
- After a node reboot (kernel upgrade), pods on
node-4.internalcannot communicate with pods on other nodes. - Pods on the node can reach the Kubernetes API server but not other pod IPs.
- Newly created pods on this node get stuck in
ContainerCreatingwith a CNI error. - Existing pods that survived the reboot show
Runningbut their network connections time out. - The cluster uses Calico CNI with VXLAN encapsulation.
- Other nodes in the cluster are unaffected.
- kubelet logs show repeated CNI plugin errors.