Merge pull request #17315 from ZJU-SEL/ubuntu-clean-kube-down

cluster/ubuntu/util.sh

@@ -515,6 +515,20 @@ function provision-masterandnode() {
   }    
 }
 
+# check whether kubelet has torn down all of the pods
+function check-pods-torn-down() {
+  local kubectl="${KUBE_ROOT}/cluster/kubectl.sh"
+  local attempt=0
+  while [[ ! -z "$(kubectl get pods | tail -n +2)" ]]; do
+    if (( attempt > 120 )); then
+      echo "timeout waiting for tearing down pods" >> ~/kube/err.log
+    fi
+    echo "waiting for tearing down pods"
+    attempt=$((attempt+1))
+    sleep 5
+  done
+}
+
 # Delete a kubernetes cluster
 function kube-down() {
   
@@ -523,6 +537,7 @@ function kube-down() {
   source "${KUBE_ROOT}/cluster/common.sh"
 
   tear_down_alive_resources
+  check-pods-torn-down 
   
   local ii=0
   for i in ${nodes}; do
@@ -543,12 +558,18 @@ function kube-down() {
             rm -rf /srv/kubernetes
             '
         " || echo "Cleaning on master ${i#*@} failed"
+
+        if [[ "${roles[${ii}]}" == "ai" ]]; then
+          ssh $SSH_OPTS -t "$i" "sudo rm -rf /var/lib/kubelet"
+        fi
+        
       elif [[ "${roles[${ii}]}" == "i" ]]; then
         echo "Cleaning on node ${i#*@}"
         ssh $SSH_OPTS -t "$i" "
           pgrep flanneld && \
           sudo -p '[sudo] password to stop node: ' -- /bin/bash -c '
             service flanneld stop
+            rm -rf /var/lib/kubelet            
             '
           " || echo "Cleaning on node ${i#*@} failed"
       else
@@ -567,10 +588,8 @@ function kube-down() {
           /etc/default/flanneld
         
         rm -rf ~/kube
-        rm -rf /var/lib/kubelet 
         rm -f /run/flannel/subnet.env
       '" || echo "cleaning legacy files on ${i#*@} failed"
-
     ((ii=ii+1))
   done
 }