Question

You deploy a Service of type LoadBalancer named 'web-lb' in an on-premises bare-metal cluster (installed via kubeadm, no cloud provider integration). After applying the manifest, running 'kubectl get svc web-lb' shows the EXTERNAL-IP column stuck in '<pending>' indefinitely, though the Service has a valid ClusterIP and healthy endpoints. What is the most likely cause and appropriate remediation?

Accepted Answer

The cluster has no external load balancer provisioner; install a solution such as MetalLB (or another bare-metal LB controller) to allocate external IPs.

Answer

The Service selector does not match any Pods; correct the selector so endpoints are populated and the external IP will be assigned.

Answer

kube-proxy is running in IPVS mode, which does not support LoadBalancer Services; switch kube-proxy to iptables mode.

Answer

The nodePort range is exhausted; expand --service-node-port-range on the API server to free an external IP.