Add random actor placement; fix cancellation callback; update test skips (#11684)

python/ray/test_utils.py

@@ -439,3 +439,7 @@ def format_web_url(url):
     if not url.startswith("http://"):
         return "http://" + url
     return url
+
+
+def new_scheduler_enabled():
+    return os.environ.get("RAY_ENABLE_NEW_SCHEDULER") == "1"

python/ray/tests/BUILD

@@ -11,7 +11,13 @@ SRCS = [] + select({
 py_test_module_list(
   files = [
     "test_async.py",
+    "test_actor.py",
+    "test_actor_failures.py",
+    "test_actor_advanced.py",
+    "test_advanced.py",
+    "test_advanced_2.py",
     "test_basic.py",
+    "test_basic_2.py",
     "test_cli.py",
     "test_component_failures_3.py",
     "test_error_ray_not_initialized.py",
@@ -27,15 +33,9 @@ py_test_module_list(
 
 py_test_module_list(
   files = [
-    "test_actor_advanced.py",
-    "test_actor_failures.py",
-    "test_actor.py",
     "test_actor_resources.py",
-    "test_advanced_2.py",
     "test_advanced_3.py",
-    "test_advanced.py",
     "test_array.py",
-    "test_basic_2.py",
     "test_cancel.py",
     "test_component_failures_2.py",
     "test_dynres.py",

python/ray/tests/test_actor_advanced.py

@@ -12,7 +12,7 @@ import ray
 import ray.test_utils
 import ray.cluster_utils
 from ray.test_utils import (run_string_as_driver, get_non_head_nodes,
-                            wait_for_condition)
+                            wait_for_condition, new_scheduler_enabled)
 from ray.experimental.internal_kv import _internal_kv_get, _internal_kv_put
 
 
@@ -91,6 +91,7 @@ def test_actor_load_balancing(ray_start_cluster):
     ray.get(results)
 
 
+@pytest.mark.skipif(new_scheduler_enabled(), reason="multi node broken")
 def test_actor_lifetime_load_balancing(ray_start_cluster):
     cluster = ray_start_cluster
     cluster.add_node(num_cpus=0)
@@ -943,6 +944,7 @@ def test_actor_creation_task_crash(ray_start_regular):
         }
     }],
     indirect=True)
+@pytest.mark.skipif(new_scheduler_enabled(), reason="todo hangs")
 def test_pending_actor_removed_by_owner(ray_start_regular):
     # Verify when an owner of pending actors is killed, the actor resources
     # are correctly returned.

python/ray/tests/test_actor_failures.py

@@ -14,6 +14,7 @@ from ray.test_utils import (
     wait_for_pid_to_exit,
     generate_system_config_map,
     get_other_nodes,
+    new_scheduler_enabled,
     SignalActor,
 )
 
@@ -265,6 +266,7 @@ def test_actor_restart_on_node_failure(ray_start_cluster):
     assert result == 1 or result == results[-1] + 1
 
 
+@pytest.mark.skipif(new_scheduler_enabled(), reason="dynamic resources todo")
 def test_actor_restart_without_task(ray_start_regular):
     """Test a dead actor can be restarted without sending task to it."""
 
@@ -483,6 +485,7 @@ def test_decorated_method(ray_start_regular):
         "num_cpus": 1,
         "num_nodes": 3,
     }], indirect=True)
+@pytest.mark.skipif(new_scheduler_enabled(), reason="dynamic resources todo")
 def test_ray_wait_dead_actor(ray_start_cluster):
     """Tests that methods completed by dead actors are returned as ready"""
     cluster = ray_start_cluster

python/ray/tests/test_advanced_2.py

@@ -14,6 +14,7 @@ import ray.test_utils
 from ray.test_utils import (
     RayTestTimeoutException,
     wait_for_condition,
+    new_scheduler_enabled,
 )
 
 logger = logging.getLogger(__name__)
@@ -251,6 +252,7 @@ def test_zero_cpus(shutdown_only):
     ray.get(x)
 
 
+@pytest.mark.skipif(new_scheduler_enabled(), reason="zero cpu handling")
 def test_zero_cpus_actor(ray_start_cluster):
     cluster = ray_start_cluster
     cluster.add_node(num_cpus=0)