Propagation of last valid node metrics instead of the [-2]nd value

Garbage colletor modification: limiting the batch size

circle/vm/models/node.py

@@ -341,15 +341,21 @@ class Node(OperatedMixin, TimeStampedModel):
                 # Example:
                 # {"target": "circle.szianode.cpu.usage",
                 #  "datapoints": [[0.6, 1403045700], [0.5, 1403045760]
+                logger.info('MONITOR_TARGET: %s', target)
                 try:
                     metric = target['target']
                     if metric.startswith(prefix):
                         metric = metric[len(prefix):]
                     else:
+            	        logger.info('MONITOR_MET: %s %s', target, metric)
                         continue
+                    value = target['datapoints'][-1][0]
+                    if value is None: 
                         value = target['datapoints'][-2][0]
                     retval[metric] = float(value)
+                    logger.info('MONITOR_RETVAL: %s %s, %s', target['target'], metric, retval[metric])
                 except (KeyError, IndexError, ValueError, TypeError):
+                    logger.info('MONITOR_ERR: %s %s', metric, value)
                     continue
 
             return retval
@@ -416,7 +422,7 @@ class Node(OperatedMixin, TimeStampedModel):
         vm_state_changed hook.
         """
         domains = {}
-        domain_list = self.remote_query(vm_tasks.list_domains_info, timeout=5,
+        domain_list = self.remote_query(vm_tasks.list_domains_info, timeout=10,
                                         priority="fast")
         if domain_list is None:
             logger.info("Monitoring failed at: %s", self.name)

circle/vm/tasks/local_periodic_tasks.py

@@ -33,7 +33,7 @@ def update_domain_states():
 
 
 @celery.task(ignore_result=True)
-def garbage_collector(timeout=15):
+def garbage_collector(offset=timezone.timedelta(seconds=20)):
     """Garbage collector for instances.
 
     Suspends and destroys expired instances.
@@ -42,8 +42,10 @@ def garbage_collector(timeout=15):
     :type timeout: int
     """
     now = timezone.now()
+    bw = 0
     for i in Instance.objects.filter(destroyed_at=None).all():
-        if i.time_of_delete and now > i.time_of_delete:
+        if i.time_of_delete and now > i.time_of_delete + offset and bw < 20:
+	    bw = bw + 1
             i.destroy.async(system=True)
             logger.info("Expired instance %d destroyed.", i.pk)
             try:
@@ -57,7 +59,8 @@ def garbage_collector(timeout=15):
                 logger.debug('Could not notify owner of instance %d .%s',
                              i.pk, unicode(e))
         elif (i.time_of_suspend and now > i.time_of_suspend and
-              i.state == 'RUNNING'):
+              i.state == 'RUNNING' and bw < 20):
+            bw = bw + 1
             logger.info("Expired instance %d suspended." % i.pk)
             try:
             	i.sleep.async(system=True)
@@ -77,4 +80,4 @@ def garbage_collector(timeout=15):
             logger.debug("Instance %d expires soon." % i.pk)
             i.notify_owners_about_expiration()
         else:
-            logger.debug("Instance %d didn't expire." % i.pk)
+            logger.debug("Instance %d didn't expire. bw:%d", i.pk, bw)