Improve node_ids_by_cluster() by adding filters

This adds a 'filters' parameter to the node_ids_by_cluster() DB API and
also exposes it through the versioned Node object. It will be used by
upper layers to save DB access overhead.

Change-Id: I412f8e5ff0e6e9ef5e7380423caeafa9d67ff8eb

senlin/db/api.py

@@ -104,8 +104,8 @@ def node_get_all_by_cluster(context, cluster_id, filters=None,
                                         project_safe=project_safe)
 
 
-def node_ids_by_cluster(context, cluster_id):
-    return IMPL.node_ids_by_cluster(context, cluster_id)
+def node_ids_by_cluster(context, cluster_id, filters=None):
+    return IMPL.node_ids_by_cluster(context, cluster_id, filters=None)
 
 
 def node_count_by_cluster(context, cluster_id, **kwargs):

senlin/db/sqlalchemy/api.py

@@ -283,12 +283,14 @@ def node_get_all_by_cluster(context, cluster_id, filters=None,
     return query.all()
 
 
-def node_ids_by_cluster(context, cluster_id):
+def node_ids_by_cluster(context, cluster_id, filters=None):
     """an internal API for getting node IDs."""
     with session_for_read() as session:
-        nodes = session.query(models.Node.id).filter_by(
-            cluster_id=cluster_id).all()
-        return [n[0] for n in nodes]
+        query = session.query(models.Node.id).filter_by(cluster_id=cluster_id)
+        if filters:
+            query = utils.exact_filter(query, models.Node, filters)
+
+        return [n[0] for n in query.all()]
 
 
 def node_count_by_cluster(context, cluster_id, **kwargs):

senlin/objects/node.py

@@ -136,6 +136,11 @@ class Node(base.SenlinObject, base.VersionedObjectDictCompat):
             context, cluster_id, filters=filters, project_safe=project_safe)
         return [cls._from_db_object(context, cls(), obj) for obj in objs]
 
+    @classmethod
+    def ids_by_cluster(cls, context, cluster_id, filters=None):
+        """An internal API for retrieving node ids only."""
+        return db_api.node_ids_by_cluster(context, cluster_id, filters=filters)
+
     @classmethod
     def count_by_cluster(cls, context, cluster_id, **kwargs):
         return db_api.node_count_by_cluster(context, cluster_id, **kwargs)

senlin/tests/unit/db/test_node_api.py

@@ -498,7 +498,7 @@ class DBAPINodeTest(base.SenlinTestCase):
                                            project_safe=False)
         self.assertEqual(2, res)
 
-    def test_nodeids_by_cluster(self):
+    def test_ids_by_cluster(self):
         node0 = shared.create_node(self.ctx, None, self.profile)
         node1 = shared.create_node(self.ctx, self.cluster, self.profile)
         node2 = shared.create_node(self.ctx, self.cluster, self.profile)
@@ -512,6 +512,23 @@ class DBAPINodeTest(base.SenlinTestCase):
         self.assertEqual(1, len(results))
         self.assertEqual(node0.id, results[0])
 
+    def test_ids_by_cluster_with_filters(self):
+        node0 = shared.create_node(self.ctx, None, self.profile,
+                                   role='slave')
+        node1 = shared.create_node(self.ctx, self.cluster, self.profile,
+                                   role='master')
+        shared.create_node(self.ctx, self.cluster, self.profile)
+
+        results = db_api.node_ids_by_cluster(self.ctx, self.cluster.id,
+                                             filters={'role': 'master'})
+        self.assertEqual(1, len(results))
+        self.assertEqual(node1.id, results[0])
+
+        # retrieve orphan nodes
+        results = db_api.node_ids_by_cluster(self.ctx, '')
+        self.assertEqual(1, len(results))
+        self.assertEqual(node0.id, results[0])
+
     def test_node_update(self):
         node = shared.create_node(self.ctx, self.cluster, self.profile)
         new_attributes = {