Kubernetes cluster with 'terminating' pods

Good evening. Can you help me. I'm using **libcluster** in Kubernetes with the strategy **Cluster.Strategy.Kubernetes** and **libring** for dynamic messaging
```
config :libcluster,
topologies: [
feed: [
strategy: Cluster.Strategy.Kubernetes,
config: [kubernetes_selector: "app=feed", kubernetes_node_basename: "feed"]
],
fixture: [
strategy: Cluster.Strategy.Kubernetes,
config: [kubernetes_selector: "app=fixture", kubernetes_node_basename: "fixture"]
],
web: [
strategy: Cluster.Strategy.Kubernetes,
config: [kubernetes_selector: "app=web", kubernetes_node_basename: "web"]
]
]
```

For some reason, sometimes after node moved to **Terminating** state, it still in cluster. Eventually, request to such node results in request timeout.