Cluster information:
Kubernetes version: 1.32
Cloud being used: bare-metal
Installation method: kubeadm
Host OS: RHEL8
CNI and version:
CRI and version:
The Kubernetes API server on a node experienced a restart, which was followed by issues with the CNI master. These problems led to network connectivity issues on several worker nodes, causing applications running on those nodes to disconnect and experience service interruptions.
Logs at below:
| etcd on master01 shows read delays (~1.6–2.0s), Raft agreement delays, peer timeouts. |
|---|
| etcd logs: ‘ignored out-of-date read index’ → leader sync problem; possible leader change starts. |
| API server on master02 logs: handler timeouts for lease updates (controller-manager/scheduler). |
| API server on master02 logs: etcdserver: request timed out, possibly due to previous leader failure. |
| Multiple kube-apiserver requests fail; CNI master on master02 loses connectivity (depends on API). |
| API server logs first “ClusterIP allocator updated” — control plane responding again. |
| Worker node connectivity gradually returns; CNI network fabric stabilizes. |