Fixed an issue occasionally causing an empty config for etcd slave nodes

The script which adds new nodes to etcd cluster now does up to 30 attempts
and fails (returns a non-zero exit code) if none are successful.

A minor performance optimization has been done as well, so no unneeded
waits happen when the command is successful and no temporary files are used.

Closes-Bug: #1494997
Change-Id: Ic0552f388518c119925da902b64797bbf96f979f

Kubernetes/KubernetesCluster/package/Resources/scripts/master-add-member.sh

@@ -2,13 +2,15 @@
 
 # $1 - NAME
 # $2 - IP
-count=5
+count=30
-done=false
 
-while  [ $count -gt 0 ] && [ "$done" != "true" ]
+while [ $count -gt 0 ]; do
-do
+ out=$(/opt/bin/etcdctl member add "$1" "http://$2:7001")
-  /opt/bin/etcdctl member add $1 http://$2:7001  > /tmp/out && done="true"
+ if [ $? -eq 0 ]; then
-  ((count-- ))
+   echo "$out" | grep ETCD_INITIAL_CLUSTER= | cut -f 2 -d '"'
-  sleep 2
+   exit 0
+ fi
+ ((count-- ))
+ sleep 2
 done
-cat /tmp/out | grep ETCD_INITIAL_CLUSTER | grep -n 1 | cut -f 2 -d'"'
+exit 1