Caching floating ips as server list is done

There are individual requets for each node to be built and
deleted to query for floating ips list. That means one call
to the nova API for each process, causing high load into the
system. Adding a cache layer that only queries for the list
every 5 seconds, as well as we are doing for the server list.

Change-Id: I43dd4805debbdf790c3ec28db686459138d59d4d

nodepool/provider_manager.py

@@ -33,6 +33,7 @@ from task_manager import Task, TaskManager, ManagerStoppedException
 
 
 SERVER_LIST_AGE = 5   # How long to keep a cached copy of the server list
+IPS_LIST_AGE = 5      # How long to keep a cached copy of the ip list
 ITERATE_INTERVAL = 2  # How long to sleep while waiting for something
                       # in a loop
 
@@ -314,6 +315,9 @@ class ProviderManager(TaskManager):
         self._servers = []
         self._servers_time = 0
         self._servers_lock = threading.Lock()
+        self._ips = []
+        self._ips_time = 0
+        self._ips_lock = threading.Lock()
 
     @property
     def _flavors(self):
@@ -541,7 +545,14 @@ class ProviderManager(TaskManager):
         return self.submitTask(ListFlavorsTask())
 
     def listFloatingIPs(self):
-        return self.submitTask(ListFloatingIPsTask())
+        if time.time() - self._ips_time >= IPS_LIST_AGE:
+            if self._ips_lock.acquire(False):
+                try:
+                    self._ips = self.submitTask(ListFloatingIPsTask())
+                    self._ips_time = time.time()
+                finally:
+                    self._ips_lock.release()
+        return self._ips
 
     def removeFloatingIP(self, server_id, address):
         return self.submitTask(RemoveFloatingIPTask(server=server_id,