neutron client needs to change polling strategy

[richbrowne]

I don't know why I closed the issue where CLUDS test was failing. But this is a request for change in the testing client.

This issue is likely caused by the way in which the polling client is initialized and implemented. In some cases the .4 second polling interval is too short and causes the api call to happen when the prior one is still in progress. This results in an exception which is not handled by the polling client.

The call to delete should not be retried. In fact, the proper way is to test the existence of the object, and if it exists try to delete, then poll for the existence until it is removed.

So the pattern is:

if x.exists()
x.delete()
while x.exists() and retries < max_retries:
sleep (interval)

if x.exists():
return ERROR