From 569eee5e71311849004d6d526bc62d8df912faf3 Mon Sep 17 00:00:00 2001 From: Eric Liang Date: Fri, 27 Nov 2020 16:10:38 -0800 Subject: [PATCH] Enable more new scheduler tests (#12421) --- python/ray/tests/test_gcs_fault_tolerance.py | 5 +---- python/ray/tests/test_multi_node.py | 4 +--- python/ray/tests/test_multi_node_2.py | 2 +- python/ray/tests/test_multi_tenancy.py | 1 - src/ray/raylet/scheduling/cluster_task_manager.h | 1 - 5 files changed, 3 insertions(+), 10 deletions(-) diff --git a/python/ray/tests/test_gcs_fault_tolerance.py b/python/ray/tests/test_gcs_fault_tolerance.py index ca802d53f..a92010bd5 100644 --- a/python/ray/tests/test_gcs_fault_tolerance.py +++ b/python/ray/tests/test_gcs_fault_tolerance.py @@ -21,7 +21,7 @@ def increase(x): return x + 1 -@pytest.mark.skipif(new_scheduler_enabled(), reason="broken") +@pytest.mark.skipif(new_scheduler_enabled(), reason="notimpl") @pytest.mark.parametrize( "ray_start_regular", [ generate_system_config_map( @@ -47,7 +47,6 @@ def test_gcs_server_restart(ray_start_regular): assert result == 2 -@pytest.mark.skipif(new_scheduler_enabled(), reason="broken") @pytest.mark.parametrize( "ray_start_regular", [ generate_system_config_map( @@ -71,7 +70,6 @@ def test_gcs_server_restart_during_actor_creation(ray_start_regular): assert len(unready) == 0 -@pytest.mark.skipif(new_scheduler_enabled(), reason="broken") @pytest.mark.parametrize( "ray_start_cluster_head", [ generate_system_config_map( @@ -132,7 +130,6 @@ def test_node_failure_detector_when_gcs_server_restart(ray_start_cluster_head): wait_for_condition(condition, timeout=10) -@pytest.mark.skipif(new_scheduler_enabled(), reason="broken") @pytest.mark.parametrize( "ray_start_regular", [ generate_system_config_map( diff --git a/python/ray/tests/test_multi_node.py b/python/ray/tests/test_multi_node.py index 582193266..cb206112d 100644 --- a/python/ray/tests/test_multi_node.py +++ b/python/ray/tests/test_multi_node.py @@ -9,7 +9,7 @@ from ray.test_utils import ( RayTestTimeoutException, check_call_ray, run_string_as_driver, run_string_as_driver_nonblocking, wait_for_children_of_pid, wait_for_children_of_pid_to_exit, wait_for_condition, kill_process_by_name, - Semaphore, init_error_pubsub, get_error_message, new_scheduler_enabled) + Semaphore, init_error_pubsub, get_error_message) def test_remote_raylet_cleanup(ray_start_cluster): @@ -139,7 +139,6 @@ print("success") assert "success" in out -@pytest.mark.skipif(new_scheduler_enabled(), reason="hangs") def test_driver_exiting_quickly(call_ray_start): # This test will create some drivers that submit some tasks and then # exit without waiting for the tasks to complete. @@ -305,7 +304,6 @@ ray.get([a.log.remote(), f.remote()]) "--min-worker-port=0 --max-worker-port=0 --port 0" ], indirect=True) -@pytest.mark.skipif(new_scheduler_enabled(), reason="hangs") def test_drivers_release_resources(call_ray_start): address = call_ray_start diff --git a/python/ray/tests/test_multi_node_2.py b/python/ray/tests/test_multi_node_2.py index 0579a0c41..04d57d280 100644 --- a/python/ray/tests/test_multi_node_2.py +++ b/python/ray/tests/test_multi_node_2.py @@ -118,7 +118,7 @@ def verify_load_metrics(monitor, expected_resource_usage=None, timeout=30): "num_cpus": 2, }], indirect=True) -@pytest.mark.skipif(new_scheduler_enabled(), reason="fails") +@pytest.mark.skipif(new_scheduler_enabled(), reason="fails in travis?") def test_heartbeats_single(ray_start_cluster_head): """Unit test for `Cluster.wait_for_nodes`. diff --git a/python/ray/tests/test_multi_tenancy.py b/python/ray/tests/test_multi_tenancy.py index c3e440235..b7caa8c08 100644 --- a/python/ray/tests/test_multi_tenancy.py +++ b/python/ray/tests/test_multi_tenancy.py @@ -251,7 +251,6 @@ ray.shutdown() assert worker2.pid == get_workers()[0].pid -@pytest.mark.skipif(new_scheduler_enabled(), reason="raylet hang 100% cpu") def test_worker_registration_failure_after_driver_exit(shutdown_only): info = ray.init(num_cpus=1) diff --git a/src/ray/raylet/scheduling/cluster_task_manager.h b/src/ray/raylet/scheduling/cluster_task_manager.h index b97dc7659..c5dba73be 100644 --- a/src/ray/raylet/scheduling/cluster_task_manager.h +++ b/src/ray/raylet/scheduling/cluster_task_manager.h @@ -128,7 +128,6 @@ class ClusterTaskManager { std::function is_owner_alive_; NodeInfoGetter get_node_info_; - // TODO (Alex): Implement fair queuing for these queues /// Queue of lease requests that are waiting for resources to become available. std::unordered_map> tasks_to_schedule_;