Our k8s cluster uses VMware VM as worker nodes. K8s version is 1.22.9, network is flannel, kube-proxy mode is iptables.
A worker node with storage was moved with vMotion. After the migration is complete, the pod ip on the worker node cannot be accessed. 50% of TCP connections to nodeport 30800 of this worker node cannot be established. No response was received after SYN packets were sent.
The nodeport svc is ingress-nginx, nodeport is 30800.I cannot access port 30800 of this worker node, but access 30800 of other worker nodes is OK.
The malfunction lasted for about 1 hour and automatically recovered without any intervention.
I suspect that vmotion may have caused the malfunction, but we don't know the exact cause.
The ping test network was almost not interrupted during the vMotion process, and ports 22 10050 of the VM could be accessed normally after the migration. I suspect that the migration caused problems in kube-proxy or flannel.
I tested it in a test environment, but it was not repeated.