[Placement Group]Add strict spread strategy (#10174)

* support STRICT_SPREAD strategy

* fix review comments

* rebase master

* fix lint error

* fix lint error

Co-authored-by: 灵洵 <fengbin.ffb@antfin.com>

python/ray/_raylet.pyx

@@ -70,6 +70,7 @@ from ray.includes.common cimport (
     PLACEMENT_STRATEGY_PACK,
     PLACEMENT_STRATEGY_SPREAD,
     PLACEMENT_STRATEGY_STRICT_PACK,
+    PLACEMENT_STRATEGY_STRICT_SPREAD,
 )
 from ray.includes.unique_ids cimport (
     CActorID,
@@ -1064,9 +1065,11 @@ cdef class CoreWorker:
             c_strategy = PLACEMENT_STRATEGY_PACK
         elif strategy == b"SPREAD":
             c_strategy = PLACEMENT_STRATEGY_SPREAD
+        elif strategy == b"STRICT_PACK":
+            c_strategy = PLACEMENT_STRATEGY_STRICT_PACK
         else:
-            if strategy == b"STRICT_PACK":
-                c_strategy = PLACEMENT_STRATEGY_STRICT_PACK
+            if strategy == b"STRICT_SPREAD":
+                c_strategy = PLACEMENT_STRATEGY_STRICT_SPREAD
             else:
                 raise TypeError(strategy)
 

python/ray/experimental/placement_group.py

@@ -19,6 +19,7 @@ def placement_group(bundles: List[Dict[str, float]],
             PACK: Packs Bundles into as few nodes as possible.
             SPREAD: Places Bundles across distinct nodes as even as possible.
             STRICT_PACK: Packs Bundles into one node.
+            STRICT_SPREAD: Packs Bundles across distinct nodes.
             The group is not allowed to span multiple nodes.
         name: The name of the placement group.
     """

python/ray/includes/common.pxd

@@ -176,6 +176,8 @@ cdef extern from "src/ray/protobuf/common.pb.h" nogil:
         "ray::PlacementStrategy::SPREAD"
     cdef CPlacementStrategy PLACEMENT_STRATEGY_STRICT_PACK \
         "ray::PlacementStrategy::STRICT_PACK"
+    cdef CPlacementStrategy PLACEMENT_STRATEGY_STRICT_SPREAD \
+        "ray::PlacementStrategy::STRICT_SPREAD"
 
 cdef extern from "ray/common/task/scheduling_resources.h" nogil:
     cdef cppclass ResourceSet "ray::ResourceSet":

python/ray/tests/test_placement_group.py

@@ -137,7 +137,7 @@ def test_placement_group_spread(ray_start_cluster):
     # Get all actors.
     actor_infos = ray.actors()
 
-    # Make sure all actors in counter_list are collocated in one node.
+    # Make sure all actors in counter_list are located in separate nodes.
     actor_info_1 = actor_infos.get(actor_1._actor_id.hex())
     actor_info_2 = actor_infos.get(actor_2._actor_id.hex())
 
@@ -148,6 +148,63 @@ def test_placement_group_spread(ray_start_cluster):
     assert node_of_actor_1 != node_of_actor_2
 
 
+def test_placement_group_strict_spread(ray_start_cluster):
+    @ray.remote(num_cpus=2)
+    class Actor(object):
+        def __init__(self):
+            self.n = 0
+
+        def value(self):
+            return self.n
+
+    cluster = ray_start_cluster
+    num_nodes = 3
+    for _ in range(num_nodes):
+        cluster.add_node(num_cpus=4)
+    ray.init(address=cluster.address)
+
+    placement_group_id = ray.experimental.placement_group(
+        name="name",
+        strategy="STRICT_SPREAD",
+        bundles=[{
+            "CPU": 2
+        }, {
+            "CPU": 2
+        }, {
+            "CPU": 2
+        }])
+    actor_1 = Actor.options(
+        placement_group_id=placement_group_id,
+        placement_group_bundle_index=0).remote()
+    actor_2 = Actor.options(
+        placement_group_id=placement_group_id,
+        placement_group_bundle_index=1).remote()
+    actor_3 = Actor.options(
+        placement_group_id=placement_group_id,
+        placement_group_bundle_index=2).remote()
+
+    print(ray.get(actor_1.value.remote()))
+    print(ray.get(actor_2.value.remote()))
+    print(ray.get(actor_3.value.remote()))
+
+    # Get all actors.
+    actor_infos = ray.actors()
+
+    # Make sure all actors in counter_list are located in separate nodes.
+    actor_info_1 = actor_infos.get(actor_1._actor_id.hex())
+    actor_info_2 = actor_infos.get(actor_2._actor_id.hex())
+    actor_info_3 = actor_infos.get(actor_3._actor_id.hex())
+
+    assert actor_info_1 and actor_info_2 and actor_info_3
+
+    node_of_actor_1 = actor_info_1["Address"]["NodeID"]
+    node_of_actor_2 = actor_info_2["Address"]["NodeID"]
+    node_of_actor_3 = actor_info_3["Address"]["NodeID"]
+    assert node_of_actor_1 != node_of_actor_2
+    assert node_of_actor_1 != node_of_actor_3
+    assert node_of_actor_2 != node_of_actor_3
+
+
 def test_placement_group_actor_resource_ids(ray_start_cluster):
     @ray.remote(num_cpus=1)
     class F: