[autoscaler] hotfix calculate_node_resources (#10874)

python/ray/autoscaler/resource_demand_scheduler.py

@@ -56,6 +56,7 @@ class ResourceDemandScheduler:
             nodes: List of existing nodes in the cluster.
             pending_nodes: Summary of node types currently being launched.
             resource_demands: Vector of resource demands from the scheduler.
+            usage_by_ip: Mapping from ip to available resources.
         """
 
         if resource_demands is None:
@@ -113,7 +114,6 @@ class ResourceDemandScheduler:
             tags = self.provider.node_tags(node_id)
             if TAG_RAY_USER_NODE_TYPE in tags:
                 node_type = tags[TAG_RAY_USER_NODE_TYPE]
-                node_type_counts[node_type] += 1
                 ip = self.provider.internal_ip(node_id)
                 available_resources = usage_by_ip.get(ip)
                 add_node(node_type, available_resources)

python/ray/tests/test_resource_demand_scheduler.py

@@ -180,6 +180,27 @@ def test_get_nodes_to_launch_limits():
     assert to_launch == []
 
 
+def test_calculate_node_resources():
+    provider = MockProvider()
+    scheduler = ResourceDemandScheduler(provider, TYPES_A, 10)
+
+    provider.create_node({}, {TAG_RAY_USER_NODE_TYPE: "p2.8xlarge"}, 2)
+
+    nodes = provider.non_terminated_nodes({})
+
+    ips = provider.non_terminated_node_ips({})
+    # 2 free p2.8xls
+    utilizations = {ip: {"GPU": 8} for ip in ips}
+    # 1 more on the way
+    pending_nodes = {"p2.8xlarge": 1}
+    # requires 4 p2.8xls (only 3 are in cluster/pending)
+    demands = [{"GPU": 8}] * (len(utilizations) + 2)
+    to_launch = scheduler.get_nodes_to_launch(nodes, pending_nodes, demands,
+                                              utilizations)
+
+    assert to_launch == [("p2.8xlarge", 1)]
+
+
 class LoadMetricsTest(unittest.TestCase):
     def testResourceDemandVector(self):
         lm = LoadMetrics()