Cluster provision and benchmark stalls using Rally

@palani

Can you add debug logging? Just "level": "INFO" to "level": "DEBUG" in your ~/.rally/logging.json, rerun and share ~/.rally/logs/rally.log.

One other test, to verify connectivity:
Ensure all rally/rallyd processes are killed on all vms.

Then on target vms vm-2 and vm-4 run:

nc -k -v -l 1900

On the coordinator vm-1 run:

printf "message from coordinator to vm-2\n" | nc 10.156.129.121 1900
printf "message from coordinator to vn-4\n" | nc 10.156.129.169 1900

You should see both messages from the coordinator on the target nodes.