fix statistics query in postgres

when aggregate functions present, postgres requires all
non-aggregate columns selected to be included in group by clause
to ensure a single value is selected.

Change-Id: Iccbcd8046cca872e8eba472f48fa392114f34426
Closes-Bug: #1256318

ceilometer/storage/impl_sqlalchemy.py

@@ -669,7 +669,8 @@ class Connection(base.Connection):
             select.extend(group_attributes)
 
         query = session.query(*select).filter(
-            models.Meter.id == models.Sample.meter_id)
+            models.Meter.id == models.Sample.meter_id)\
+            .group_by(models.Meter.unit)
 
         if groupby:
             query = query.group_by(*group_attributes)