I have a local(without cloud provider) cluster made up of 3 vm the master and the nodes, I have created a volume with a nfs to reuse it if a pod die and is reschedule on another nodes, but i think same component not work well: I use to create the cluster just this guide: kubernetes guide and I have after that create the cluster this is the actual state:

master@master-VirtualBox:~/Documents/KubeT/nfs$ sudo kubectl get pod --all-namespaces 
    [sudo] password for master: 
    NAMESPACE     NAME                                        READY     STATUS    RESTARTS   AGE
    default       mysqlnfs3                                   1/1       Running   0          27m
    kube-system   etcd-master-virtualbox                      1/1       Running   0          46m
    kube-system   kube-apiserver-master-virtualbox            1/1       Running   0          46m
    kube-system   kube-controller-manager-master-virtualbox   1/1       Running   0          46m
    kube-system   kube-dns-86f4d74b45-f6hpf                   3/3       Running   0          47m
    kube-system   kube-flannel-ds-nffv6                       1/1       Running   0          38m
    kube-system   kube-flannel-ds-rqw9v                       1/1       Running   0          39m
    kube-system   kube-flannel-ds-s5wzn                       1/1       Running   0          44m
    kube-system   kube-proxy-6j7p8                            1/1       Running   0          38m
    kube-system   kube-proxy-7pj8d                            1/1       Running   0          39m
    kube-system   kube-proxy-jqshs                            1/1       Running   0          47m
    kube-system   kube-scheduler-master-virtualbox            1/1       Running   0          46m


master@master-VirtualBox:~/Documents/KubeT/nfs$ sudo kubectl get node
    NAME                STATUS    ROLES     AGE       VERSION
    host1-virtualbox    Ready     <none>    39m       v1.10.2
    host2-virtualbox    Ready     <none>    40m       v1.10.2
    master-virtualbox   Ready     master    48m       v1.10.2

and this is the pod:

master@master-VirtualBox:~/Documents/KubeT/nfs$ sudo kubectl get pod
    NAME        READY     STATUS    RESTARTS   AGE
    mysqlnfs3   1/1       Running   0          29m

it is schedule on the host2 and if i try to go in the shell of host 2 and I do dockerexec I use the container very well, the data are store and retrieve, but when I try to use kubect exec not work:

master@master-VirtualBox:~/Documents/KubeT/nfs$ sudo kubectl exec -it -n default mysqlnfs3 -- /bin/bash
 error: unable to upgrade connection: pod does not exist