k8s: Make it work even if kubectl is not installed on the machine

There's code in the k8s provider that takes care of pulling the
kubectl binary from the cluster and use that one for subsequent
operations; however, right before that happens, a few settings
are changed in kubeconfig, which means that it's necessary for
the host system to provide its own kubectl binary.

Invert the order of these operations, and make sure that we
always invoke the binary obtained from the cluster. This way, it
is possible to successfully bring up a cluster on a machine that
doesn't have kubectl installed.

kubevirt#1293

Signed-off-by: Andrea Bolognani <[email protected]>

cluster-up/cluster/k8s-provider-common.sh

@@ -48,16 +48,15 @@ function up() {
     eval ${_cli:?} run $params
 
     ${_cli} scp --prefix $provider_prefix /etc/kubernetes/admin.conf - >${KUBEVIRTCI_CONFIG_PATH}/$KUBEVIRT_PROVIDER/.kubeconfig
-
-    # Set server and disable tls check
-    export KUBECONFIG=${KUBEVIRTCI_CONFIG_PATH}/$KUBEVIRT_PROVIDER/.kubeconfig
-    kubectl config set-cluster kubernetes --server="https://$(_main_ip):$(_port k8s)"
-    kubectl config set-cluster kubernetes --insecure-skip-tls-verify=true
-
     ${_cli} scp --prefix ${provider_prefix:?} /usr/bin/kubectl - >${KUBEVIRTCI_CONFIG_PATH}/$KUBEVIRT_PROVIDER/.kubectl
 
     chmod u+x ${KUBEVIRTCI_CONFIG_PATH}/$KUBEVIRT_PROVIDER/.kubectl
 
+    # Set server and disable tls check
+    export KUBECONFIG=${KUBEVIRTCI_CONFIG_PATH}/$KUBEVIRT_PROVIDER/.kubeconfig
+    ${KUBEVIRTCI_CONFIG_PATH}/$KUBEVIRT_PROVIDER/.kubectl config set-cluster kubernetes --server="https://$(_main_ip):$(_port k8s)"
+    ${KUBEVIRTCI_CONFIG_PATH}/$KUBEVIRT_PROVIDER/.kubectl config set-cluster kubernetes --insecure-skip-tls-verify=true
+
     # Make sure that local config is correct
     prepare_config
     ssh="${_cli} --prefix $provider_prefix ssh"