Remove conditional blocking on server list

When the server list cache proceduce was imported from nodepool,
it grew this internal conditional which caused all server list
calls to block on aquiring the server list cache update lock iff
the server list cache is empty.  This means that on startup, we
could have a thundering herd of threads which all insisted on
waiting for the lock because there was no cached server list.  This
does not seem to be necessary, and the original nodepool logic of
always performing non-blocking acquisitions of the lock and
returning the old (possibly empty) list of servers for any thread
that did not get the lock (but will try again a few seconds later
and should receive updated data at that time) should be sufficient.

Change-Id: I5640f60da2b7789a98bea033e16695389c6062e0

shade/openstackcloud.py

@@ -1150,9 +1150,7 @@ class OpenStackCloud(object):
             # a lock, and the non-blocking acquire method will cause
             # subsequent threads to just skip this and use the old
             # data until it succeeds.
-            # For the first time, when there is no data, make the call
-            # blocking.
-            if self._ports_lock.acquire(len(self._ports) == 0):
+            if self._ports_lock.acquire(False):
                 try:
                     self._ports = self._list_ports(filters)
                     self._ports_time = time.time()
@@ -1270,9 +1268,7 @@ class OpenStackCloud(object):
             # a lock, and the non-blocking acquire method will cause
             # subsequent threads to just skip this and use the old
             # data until it succeeds.
-            # For the first time, when there is no data, make the call
-            # blocking.
-            if self._servers_lock.acquire(len(self._servers) == 0):
+            if self._servers_lock.acquire(False):
                 try:
                     self._servers = self._list_servers(detailed=detailed)
                     self._servers_time = time.time()