diff --git a/.travis.yml b/.travis.yml
index 8a589ad9e..5d1667670 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -147,13 +147,13 @@ script:
   - ./ci/suppress_output bazel test --build_tests_only --show_progress_rate_limit=100 --test_output=errors //:all
 
   # ray serve tests
-  - if [ $RAY_CI_SERVE_AFFECTED == "1" ]; then python -c 'import sys;exit(sys.version_info>=(3,5))' || python -m pytest -v --durations=5 --timeout=300 python/ray/experimental/serve/tests; fi
-  - if [ $RAY_CI_SERVE_AFFECTED == "1" ]; then python -c 'import sys;exit(sys.version_info>=(3,5))' || ./ci/suppress_output python python/ray/experimental/serve/examples/echo_full.py; fi
+  - if [ $RAY_CI_SERVE_AFFECTED == "1" ]; then python -c 'import sys;exit(sys.version_info>=(3,5))' || RAY_FORCE_DIRECT=0 python -m pytest -v --durations=5 --timeout=300 python/ray/experimental/serve/tests; fi
+  - if [ $RAY_CI_SERVE_AFFECTED == "1" ]; then python -c 'import sys;exit(sys.version_info>=(3,5))' || RAY_FORCE_DIRECT=0 ./ci/suppress_output python python/ray/experimental/serve/examples/echo_full.py; fi
 
   # ray tests
   # Python3.5+ only. Otherwise we will get `SyntaxError` regardless of how we set the tester.
-  - if [ $RAY_CI_PYTHON_AFFECTED == "1" ]; then python -c 'import sys;exit(sys.version_info>=(3,5))' || python -m pytest -v --durations=5 --timeout=300 python/ray/experimental/test/async_test.py; fi
-  - if [ $RAY_CI_PYTHON_AFFECTED == "1" ]; then python -c 'import sys;exit(sys.version_info>=(3,5))' || python -m pytest -v --durations=5 --timeout=300 python/ray/tests/py3_test.py; fi
+  - if [ $RAY_CI_PYTHON_AFFECTED == "1" ]; then python -c 'import sys;exit(sys.version_info>=(3,5))' || RAY_FORCE_DIRECT=0 python -m pytest -v --durations=5 --timeout=300 python/ray/experimental/test/async_test.py; fi
+  - if [ $RAY_CI_PYTHON_AFFECTED == "1" ]; then python -c 'import sys;exit(sys.version_info>=(3,5))' || RAY_FORCE_DIRECT=0 python -m pytest -v --durations=5 --timeout=300 python/ray/tests/py3_test.py; fi
 
   # py bazel tests, run using local strategy since PY2 breaks with sandbox
   - ./ci/keep_alive bazel test --spawn_strategy=local --flaky_test_attempts=3 --python_version=$BAZEL_PYTHON_VERSION --nocache_test_results --test_verbose_timeout_warnings --incompatible_allow_python_version_transitions=false --incompatible_py3_is_default=false --progress_report_interval=100 --show_progress_rate_limit=100 --show_timestamps --test_output=errors --test_tag_filters=-jenkins_only python/ray/...
diff --git a/ci/jenkins_tests/run_rllib_tests.sh b/ci/jenkins_tests/run_rllib_tests.sh
index 05db3d3ab..593a4f6c2 100755
--- a/ci/jenkins_tests/run_rllib_tests.sh
+++ b/ci/jenkins_tests/run_rllib_tests.sh
@@ -1,270 +1,270 @@
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_catalog.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_catalog.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_optimizers.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_optimizers.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_filters.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_filters.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_evaluators.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_evaluators.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_eager_support.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_eager_support.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env PongDeterministic-v0 \
     --run A3C \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 2}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env Pong-ram-v4 \
     --run A3C \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 2}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env PongDeterministic-v0 \
     --run A2C \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 2}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v1 \
     --run PPO \
     --stop '{"training_iteration": 1}' \
     --config '{"kl_coeff": 1.0, "num_sgd_iter": 10, "lr": 1e-4, "sgd_minibatch_size": 64, "train_batch_size": 2000, "num_workers": 1, "model": {"free_log_std": true}}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v1 \
     --run PPO \
     --stop '{"training_iteration": 1}' \
     --config '{"simple_optimizer": false, "num_sgd_iter": 2, "model": {"use_lstm": true}}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v1 \
     --run PPO \
     --stop '{"training_iteration": 1}' \
     --config '{"simple_optimizer": true, "num_sgd_iter": 2, "model": {"use_lstm": true}}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v1 \
     --run PPO \
     --stop '{"training_iteration": 1}' \
     --config '{"kl_coeff": 1.0, "num_sgd_iter": 10, "lr": 1e-4, "sgd_minibatch_size": 64, "train_batch_size": 2000, "num_workers": 1, "use_gae": false, "batch_mode": "complete_episodes"}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v1 \
     --run PPO \
     --stop '{"training_iteration": 1}' \
     --config '{"remote_worker_envs": true, "remote_env_batch_wait_ms": 99999999, "num_envs_per_worker": 2, "num_workers": 1, "train_batch_size": 100, "sgd_minibatch_size": 50}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v1 \
     --run PPO \
     --stop '{"training_iteration": 2}' \
     --config '{"remote_worker_envs": true, "num_envs_per_worker": 2, "num_workers": 1, "train_batch_size": 100, "sgd_minibatch_size": 50}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env Pendulum-v0 \
     --run APPO \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 2, "num_gpus": 0}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env Pendulum-v0 \
     --run ES \
     --stop '{"training_iteration": 1}' \
     --config '{"stepsize": 0.01, "episodes_per_batch": 20, "train_batch_size": 100, "num_workers": 2}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env Pong-v0 \
     --run ES \
     --stop '{"training_iteration": 1}' \
     --config '{"stepsize": 0.01, "episodes_per_batch": 20, "train_batch_size": 100, "num_workers": 2}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run A3C \
     --stop '{"training_iteration": 1}' \
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run DQN \
     --stop '{"training_iteration": 1}' \
     --config '{"lr": 1e-3, "schedule_max_timesteps": 100000, "exploration_fraction": 0.1, "exploration_final_eps": 0.02, "dueling": false, "hiddens": [], "model": {"fcnet_hiddens": [64], "fcnet_activation": "relu"}}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run DQN \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 2}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run APEX \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 2, "timesteps_per_iteration": 1000, "num_gpus": 0, "min_iter_time_s": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env FrozenLake-v0 \
     --run DQN \
     --stop '{"training_iteration": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env FrozenLake-v0 \
     --run PPO \
     --stop '{"training_iteration": 1}' \
     --config '{"num_sgd_iter": 10, "sgd_minibatch_size": 64, "train_batch_size": 1000, "num_workers": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env PongDeterministic-v4 \
     --run DQN \
     --stop '{"training_iteration": 1}' \
     --config '{"lr": 1e-4, "schedule_max_timesteps": 2000000, "buffer_size": 10000, "exploration_fraction": 0.1, "exploration_final_eps": 0.01, "sample_batch_size": 4, "learning_starts": 10000, "target_network_update_freq": 1000, "gamma": 0.99, "prioritized_replay": true}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env MontezumaRevenge-v0 \
     --run PPO \
     --stop '{"training_iteration": 1}' \
     --config '{"kl_coeff": 1.0, "num_sgd_iter": 10, "lr": 1e-4, "sgd_minibatch_size": 64, "train_batch_size": 2000, "num_workers": 1, "model": {"dim": 40, "conv_filters": [[16, [8, 8], 4], [32, [4, 4], 2], [512, [5, 5], 1]]}}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v1 \
     --run A3C \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 2, "model": {"use_lstm": true}}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run DQN \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 2}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run PG \
     --stop '{"training_iteration": 1}' \
     --config '{"sample_batch_size": 500, "num_workers": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run PG \
     --stop '{"training_iteration": 1}' \
     --config '{"sample_batch_size": 500, "use_pytorch": true}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run PG \
     --stop '{"training_iteration": 1}' \
     --config '{"sample_batch_size": 500, "num_workers": 1, "model": {"use_lstm": true, "max_seq_len": 100}}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run PG \
     --stop '{"training_iteration": 1}' \
     --config '{"sample_batch_size": 500, "num_workers": 1, "num_envs_per_worker": 10}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env Pong-v0 \
     --run PG \
     --stop '{"training_iteration": 1}' \
     --config '{"sample_batch_size": 500, "num_workers": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env FrozenLake-v0 \
     --run PG \
     --stop '{"training_iteration": 1}' \
     --config '{"sample_batch_size": 500, "num_workers": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env Pendulum-v0 \
     --run DDPG \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run IMPALA \
     --stop '{"training_iteration": 1}' \
     --config '{"num_gpus": 0, "num_workers": 2, "min_iter_time_s": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run IMPALA \
     --stop '{"training_iteration": 1}' \
     --config '{"num_gpus": 0, "num_workers": 2, "num_aggregation_workers": 2, "min_iter_time_s": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run IMPALA \
     --stop '{"training_iteration": 1}' \
     --config '{"num_gpus": 0, "num_workers": 2, "min_iter_time_s": 1, "model": {"use_lstm": true}}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run IMPALA \
     --stop '{"training_iteration": 1}' \
     --config '{"num_gpus": 0, "num_workers": 2, "min_iter_time_s": 1, "num_data_loader_buffers": 2, "replay_buffer_num_slots": 100, "replay_proportion": 1.0}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run IMPALA \
     --stop '{"training_iteration": 1}' \
     --config '{"num_gpus": 0, "num_workers": 2, "min_iter_time_s": 1, "num_data_loader_buffers": 2, "replay_buffer_num_slots": 100, "replay_proportion": 1.0, "model": {"use_lstm": true}}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env MountainCarContinuous-v0 \
     --run DDPG \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env MountainCarContinuous-v0 \
     --run DDPG \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env Pendulum-v0 \
     --run APEX_DDPG \
     --ray-num-cpus 8 \
@@ -272,7 +272,7 @@ docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     --config '{"num_workers": 2, "optimizer": {"num_replay_buffer_shards": 1}, "learning_starts": 100, "min_iter_time_s": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env Pendulum-v0 \
     --run APEX_DDPG \
     --ray-num-cpus 8 \
@@ -280,199 +280,199 @@ docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     --config '{"num_workers": 2, "optimizer": {"num_replay_buffer_shards": 1}, "learning_starts": 100, "min_iter_time_s": 1, "batch_mode": "complete_episodes", "parameter_noise": false}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run MARWIL \
     --stop '{"training_iteration": 1}' \
     --config '{"input": "/ray/rllib/tests/data/cartpole_small", "learning_starts": 0, "input_evaluation": ["wis", "is"], "shuffle_buffer_size": 10}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v0 \
     --run DQN \
     --stop '{"training_iteration": 1}' \
     --config '{"input": "/ray/rllib/tests/data/cartpole_small", "learning_starts": 0, "input_evaluation": ["wis", "is"], "soft_q": true}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_local.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_local.py
     
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_reproducibility.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_reproducibility.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_dependency.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_dependency.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_legacy.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_legacy.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_io.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_io.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_checkpoint_restore.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_checkpoint_restore.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_rollout_worker.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_rollout_worker.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_nested_spaces.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_nested_spaces.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_external_env.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_external_env.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_external_multi_agent_env.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_external_multi_agent_env.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/custom_keras_model.py --run=A2C --stop=50
+    /ray/ci/suppress_output python /ray/rllib/examples/custom_keras_model.py --run=A2C --stop=50
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/custom_keras_model.py --run=PPO --stop=50
+    /ray/ci/suppress_output python /ray/rllib/examples/custom_keras_model.py --run=PPO --stop=50
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/custom_keras_model.py --run=DQN --stop=50
+    /ray/ci/suppress_output python /ray/rllib/examples/custom_keras_model.py --run=DQN --stop=50
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/parametric_action_cartpole.py --run=PG --stop=50
+    /ray/ci/suppress_output python /ray/rllib/examples/parametric_action_cartpole.py --run=PG --stop=50
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/parametric_action_cartpole.py --run=PPO --stop=50
+    /ray/ci/suppress_output python /ray/rllib/examples/parametric_action_cartpole.py --run=PPO --stop=50
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/parametric_action_cartpole.py --run=DQN --stop=50
+    /ray/ci/suppress_output python /ray/rllib/examples/parametric_action_cartpole.py --run=DQN --stop=50
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_lstm.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_lstm.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/batch_norm_model.py --num-iters=1 --run=PPO
+    /ray/ci/suppress_output python /ray/rllib/examples/batch_norm_model.py --num-iters=1 --run=PPO
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/batch_norm_model.py --num-iters=1 --run=PG
+    /ray/ci/suppress_output python /ray/rllib/examples/batch_norm_model.py --num-iters=1 --run=PG
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/batch_norm_model.py --num-iters=1 --run=DQN
+    /ray/ci/suppress_output python /ray/rllib/examples/batch_norm_model.py --num-iters=1 --run=DQN
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/batch_norm_model.py --num-iters=1 --run=DDPG
+    /ray/ci/suppress_output python /ray/rllib/examples/batch_norm_model.py --num-iters=1 --run=DDPG
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_multi_agent_env.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_multi_agent_env.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_supported_spaces.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_supported_spaces.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_env_with_subprocess.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_env_with_subprocess.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/tests/test_rollout.sh
+    /ray/ci/suppress_output /ray/rllib/tests/test_rollout.sh
 
 # Run all single-agent regression tests (3x retry each)
 for yaml in $(ls $ROOT_DIR/../../rllib/tuned_examples/regression_tests); do
     docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-        /ray/ci/suppress_output --force-direct python /ray/rllib/tests/run_regression_tests.py \
+        /ray/ci/suppress_output python /ray/rllib/tests/run_regression_tests.py \
             /ray/rllib/tuned_examples/regression_tests/$yaml
 done
 
 # Try a couple times since it's stochastic
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-        /ray/ci/suppress_output --force-direct python /ray/rllib/tests/multiagent_pendulum.py || \
+        /ray/ci/suppress_output python /ray/rllib/tests/multiagent_pendulum.py || \
     docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-        /ray/ci/suppress_output --force-direct python /ray/rllib/tests/multiagent_pendulum.py || \
+        /ray/ci/suppress_output python /ray/rllib/tests/multiagent_pendulum.py || \
     docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-        /ray/ci/suppress_output --force-direct python /ray/rllib/tests/multiagent_pendulum.py
+        /ray/ci/suppress_output python /ray/rllib/tests/multiagent_pendulum.py
 
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/multiagent_cartpole.py --num-iters=2
+    /ray/ci/suppress_output python /ray/rllib/examples/multiagent_cartpole.py --num-iters=2
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/multiagent_two_trainers.py --num-iters=2
+    /ray/ci/suppress_output python /ray/rllib/examples/multiagent_two_trainers.py --num-iters=2
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_avail_actions_qmix.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_avail_actions_qmix.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/cartpole_lstm.py --run=PPO --stop=200
+    /ray/ci/suppress_output python /ray/rllib/examples/cartpole_lstm.py --run=PPO --stop=200
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/cartpole_lstm.py --run=IMPALA --stop=100
+    /ray/ci/suppress_output python /ray/rllib/examples/cartpole_lstm.py --run=IMPALA --stop=100
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/cartpole_lstm.py --stop=200 --use-prev-action-reward
+    /ray/ci/suppress_output python /ray/rllib/examples/cartpole_lstm.py --stop=200 --use-prev-action-reward
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/custom_loss.py --iters=2
+    /ray/ci/suppress_output python /ray/rllib/examples/custom_loss.py --iters=2
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/rollout_worker_custom_workflow.py
+    /ray/ci/suppress_output python /ray/rllib/examples/rollout_worker_custom_workflow.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/eager_execution.py --iters=2
+    /ray/ci/suppress_output python /ray/rllib/examples/eager_execution.py --iters=2
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/custom_tf_policy.py --iters=2
+    /ray/ci/suppress_output python /ray/rllib/examples/custom_tf_policy.py --iters=2
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/custom_torch_policy.py --iters=2
+    /ray/ci/suppress_output python /ray/rllib/examples/custom_torch_policy.py --iters=2
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/rollout_worker_custom_workflow.py
+    /ray/ci/suppress_output python /ray/rllib/examples/rollout_worker_custom_workflow.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/custom_metrics_and_callbacks.py --num-iters=2
+    /ray/ci/suppress_output python /ray/rllib/examples/custom_metrics_and_callbacks.py --num-iters=2
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/contrib/random_agent/random_agent.py
+    /ray/ci/suppress_output python /ray/rllib/contrib/random_agent/random_agent.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/contrib/alpha_zero/examples/train_cartpole.py --training-iteration=1
+    /ray/ci/suppress_output python /ray/rllib/contrib/alpha_zero/examples/train_cartpole.py --training-iteration=1
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/centralized_critic.py --stop=2000
+    /ray/ci/suppress_output python /ray/rllib/examples/centralized_critic.py --stop=2000
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/centralized_critic_2.py --stop=2000
+    /ray/ci/suppress_output python /ray/rllib/examples/centralized_critic_2.py --stop=2000
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/twostep_game.py --stop=2000 --run=contrib/MADDPG
+    /ray/ci/suppress_output python /ray/rllib/examples/twostep_game.py --stop=2000 --run=contrib/MADDPG
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/twostep_game.py --stop=2000 --run=PG
+    /ray/ci/suppress_output python /ray/rllib/examples/twostep_game.py --stop=2000 --run=PG
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/twostep_game.py --stop=2000 --run=QMIX
+    /ray/ci/suppress_output python /ray/rllib/examples/twostep_game.py --stop=2000 --run=QMIX
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/twostep_game.py --stop=2000 --run=APEX_QMIX
+    /ray/ci/suppress_output python /ray/rllib/examples/twostep_game.py --stop=2000 --run=APEX_QMIX
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/autoregressive_action_dist.py --stop=150
+    /ray/ci/suppress_output python /ray/rllib/examples/autoregressive_action_dist.py --stop=150
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env PongDeterministic-v4 \
     --run A3C \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 2, "use_pytorch": true, "sample_async": false, "model": {"use_lstm": false, "grayscale": true, "zero_mean": false, "dim": 84}, "preprocessor_pref": "rllib"}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env CartPole-v1 \
     --run A3C \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 2, "use_pytorch": true, "sample_async": false}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env Pendulum-v0 \
     --run A3C \
     --stop '{"training_iteration": 1}' \
     --config '{"num_workers": 2, "use_pytorch": true, "sample_async": false}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct /ray/rllib/train.py \
+    /ray/ci/suppress_output /ray/rllib/train.py \
     --env PongDeterministic-v4 \
     --run IMPALA \
     --stop='{"timesteps_total": 40000}' \
@@ -480,13 +480,13 @@ docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     --config '{"num_workers": 1, "num_gpus": 0, "num_envs_per_worker": 32, "sample_batch_size": 50, "train_batch_size": 50, "learner_queue_size": 1}'
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/agents/impala/vtrace_test.py
+    /ray/ci/suppress_output python /ray/rllib/agents/impala/vtrace_test.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/tests/test_ignore_worker_failure.py
+    /ray/ci/suppress_output python /ray/rllib/tests/test_ignore_worker_failure.py
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/custom_keras_rnn_model.py --run=PPO --stop=50 --env=RepeatAfterMeEnv
+    /ray/ci/suppress_output python /ray/rllib/examples/custom_keras_rnn_model.py --run=PPO --stop=50 --env=RepeatAfterMeEnv
 
 docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
-    /ray/ci/suppress_output --force-direct python /ray/rllib/examples/custom_keras_rnn_model.py --run=PPO --stop=50 --env=RepeatInitialEnv
+    /ray/ci/suppress_output python /ray/rllib/examples/custom_keras_rnn_model.py --run=PPO --stop=50 --env=RepeatInitialEnv
diff --git a/ci/jenkins_tests/run_tune_tests.sh b/ci/jenkins_tests/run_tune_tests.sh
index 899a9ba78..5c63f9f7c 100755
--- a/ci/jenkins_tests/run_tune_tests.sh
+++ b/ci/jenkins_tests/run_tune_tests.sh
@@ -31,147 +31,147 @@ fi
 
 echo "Using Docker image" $DOCKER_SHA
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     pytest /ray/python/ray/tune/tests/test_cluster.py
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     pytest /ray/python/ray/tune/tests/test_actor_reuse.py
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     pytest /ray/python/ray/tune/tests/test_tune_restore.py
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/tests/example.py
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     bash -c 'pip install -U tensorflow && python /ray/python/ray/tune/tests/test_logger.py'
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     bash -c 'pip install -U tensorflow==1.15 && python /ray/python/ray/tune/tests/test_logger.py'
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     bash -c 'pip install -U tensorflow==1.14 && python /ray/python/ray/tune/tests/test_logger.py'
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     bash -c 'pip install -U tensorflow==1.12 && python /ray/python/ray/tune/tests/test_logger.py'
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} -e MPLBACKEND=Agg $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} -e MPLBACKEND=Agg $DOCKER_SHA \
     python /ray/python/ray/tune/tests/tutorial.py
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/pbt_example.py \
     --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/hyperband_example.py \
     --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/async_hyperband_example.py \
     --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/tf_mnist_example.py --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/lightgbm_example.py
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/xgboost_example.py
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/logging_example.py \
     --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/mlflow_example.py
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/bayesopt_example.py \
     --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/hyperopt_example.py \
     --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} -e SIGOPT_KEY $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} -e SIGOPT_KEY $DOCKER_SHA \
     python /ray/python/ray/tune/examples/sigopt_example.py \
     --smoke-test
 
 # Runs only on Python3
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/nevergrad_example.py \
     --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/tune_mnist_keras.py \
     --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/mnist_pytorch.py --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/mnist_pytorch_trainable.py \
     --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/genetic_example.py \
     --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/skopt_example.py \
     --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/tune/examples/pbt_memnn_example.py \
     --smoke-test
 
 # uncomment once statsmodels is updated.
-# $SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+# $SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
 #     python /ray/python/ray/tune/examples/bohb_example.py \
 #     --smoke-test
 
 
 ######################## SGD TESTS #################################
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python -m pytest /ray/python/ray/experimental/sgd/tests
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/train_example.py
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/train_example.py --num-replicas=2
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/tune_example.py
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/tune_example.py --num-replicas=2
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/cifar_pytorch_example.py --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/cifar_pytorch_example.py --smoke-test --num-replicas=2
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/cifar_pytorch_example.py --smoke-test --tune
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/tensorflow_train_example.py
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/tensorflow_train_example.py --num-replicas=2
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/tensorflow_train_example.py --tune
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/cifar_tf_example.py --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/cifar_tf_example.py --num-replicas 2 --smoke-test
 
-$SUPPRESS_OUTPUT --force-direct docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
+$SUPPRESS_OUTPUT docker run --rm --shm-size=${SHM_SIZE} --memory=${MEMORY_SIZE} $DOCKER_SHA \
     python /ray/python/ray/experimental/sgd/examples/cifar_tf_example.py --num-replicas 2 --smoke-test --augment-data
diff --git a/ci/suppress_output b/ci/suppress_output
index 8510c5367..583bd6275 100755
--- a/ci/suppress_output
+++ b/ci/suppress_output
@@ -19,12 +19,6 @@ watchdog() {
 watchdog & 2>/dev/null
 WATCHDOG_PID=$!
 
-if [ "$1" == "--force-direct" ]; then
-    echo "Setting RAY_FORCE_DIRECT=1"
-    export RAY_FORCE_DIRECT=1
-    shift
-fi
-
 time "$@" >$TMPFILE 2>&1
 
 CODE=$?
diff --git a/python/ray/ray_constants.py b/python/ray/ray_constants.py
index c0c2c39e9..adb8ea6be 100644
--- a/python/ray/ray_constants.py
+++ b/python/ray/ray_constants.py
@@ -17,7 +17,7 @@ def env_integer(key, default):
 
 
 def direct_call_enabled():
-    return bool(int(os.environ.get("RAY_FORCE_DIRECT", "0")))
+    return bool(int(os.environ.get("RAY_FORCE_DIRECT", "1")))
 
 
 ID_SIZE = 20
diff --git a/python/ray/ray_perf.py b/python/ray/ray_perf.py
index 78ab7eaa3..99ca4d886 100644
--- a/python/ray/ray_perf.py
+++ b/python/ray/ray_perf.py
@@ -151,19 +151,22 @@ def main():
 
     timeit("1:1 actor calls async", actor_async, 1000)
 
-    a = Actor.options(is_direct_call=True).remote()
+    a = Actor.options(max_concurrency=16).remote()
 
     def actor_concurrent():
         ray.get([a.small_value.remote() for _ in range(1000)])
 
-    timeit("1:1 direct actor calls async", actor_concurrent, 1000)
+    timeit("1:1 actor calls concurrent", actor_concurrent, 1000)
 
-    a = Actor.options(is_direct_call=True, max_concurrency=16).remote()
+    n = 5000
+    n_cpu = multiprocessing.cpu_count() // 2
+    actors = [Actor._remote() for _ in range(n_cpu)]
+    client = Client.remote(actors)
 
-    def actor_concurrent():
-        ray.get([a.small_value.remote() for _ in range(1000)])
+    def actor_async_direct():
+        ray.get(client.small_value_batch.remote(n))
 
-    timeit("1:1 direct actor calls concurrent", actor_concurrent, 1000)
+    timeit("1:n actor calls async", actor_async_direct, n * len(actors))
 
     n_cpu = multiprocessing.cpu_count() // 2
     a = [Actor.remote() for _ in range(n_cpu)]
@@ -177,44 +180,16 @@ def main():
 
     timeit("n:n actor calls async", actor_multi2, m * n)
 
-    n = 5000
-    n_cpu = multiprocessing.cpu_count() // 2
-    actors = [Actor._remote(is_direct_call=True) for _ in range(n_cpu)]
-    client = Client.remote(actors)
-
-    def actor_async_direct():
-        ray.get(client.small_value_batch.remote(n))
-
-    timeit("1:n direct actor calls async", actor_async_direct, n * len(actors))
-
-    clients = [Client.remote(a) for a in actors]
-
-    def actor_multi2_direct():
-        ray.get([c.small_value_batch.remote(n) for c in clients])
-
-    timeit("n:n direct actor calls async", actor_multi2_direct,
-           n * len(clients))
-
     n = 1000
-    actors = [Actor._remote(is_direct_call=True) for _ in range(n_cpu)]
+    actors = [Actor._remote() for _ in range(n_cpu)]
     clients = [Client.remote(a) for a in actors]
 
     def actor_multi2_direct_arg():
         ray.get([c.small_value_batch_arg.remote(n) for c in clients])
 
-    timeit("n:n direct actor calls with arg async", actor_multi2_direct_arg,
+    timeit("n:n actor calls with arg async", actor_multi2_direct_arg,
            n * len(clients))
 
-    n = 1000
-    actors = [Actor._remote(is_direct_call=True) for _ in range(n_cpu)]
-    clients = [Client.remote(a) for a in actors]
-
-    def actor_multi2_direct_arg():
-        ray.get([c.small_value_batch_arg.remote(n) for c in clients])
-
-    timeit("multi client direct actor calls with arg async",
-           actor_multi2_direct_arg, n * len(clients))
-
 
 if __name__ == "__main__":
     main()
diff --git a/python/ray/tests/BUILD b/python/ray/tests/BUILD
index cf7e7c847..80bbb2ba5 100644
--- a/python/ray/tests/BUILD
+++ b/python/ray/tests/BUILD
@@ -6,14 +6,6 @@ py_test(
     deps = ["//:ray_lib"],
 )
 
-py_test(
-    name = "test_actor_direct",
-    size = "medium",
-    srcs = ["test_actor_direct.py", "test_actor.py"],
-    tags = ["exclusive"],
-    deps = ["//:ray_lib"],
-)
-
 py_test(
     name = "test_actor_resources",
     size = "medium",
@@ -22,19 +14,12 @@ py_test(
     deps = ["//:ray_lib"],
 )
 
-py_test(
-    name = "test_actor_resources_direct",
-    size = "medium",
-    srcs = ["test_actor_resources_direct.py", "test_actor_resources.py"],
-    tags = ["exclusive"],
-    deps = ["//:ray_lib"],
-)
-
 py_test(
     name = "test_actor_failures",
     size = "medium",
     srcs = ["test_actor_failures.py"],
-    tags = ["exclusive"],
+    # TODO(ekl) enable this once we support actor reconstruction again
+    tags = ["exclusive", "manual"],
     deps = ["//:ray_lib"],
 )
 
@@ -46,14 +31,6 @@ py_test(
     deps = ["//:ray_lib"],
 )
 
-py_test(
-    name = "test_basic_direct",
-    size = "medium",
-    srcs = ["test_basic_direct.py", "test_basic.py"],
-    tags = ["exclusive"],
-    deps = ["//:ray_lib"],
-)
-
 py_test(
     name = "test_advanced",
     size = "medium",
@@ -126,14 +103,6 @@ py_test(
     deps = ["//:ray_lib"],
 )
 
-py_test(
-    name = "test_stress_direct",
-    size = "medium",
-    srcs = ["test_stress_direct.py", "test_stress.py"],
-    tags = ["exclusive"],
-    deps = ["//:ray_lib"],
-)
-
 py_test(
     name = "test_stress_sharded",
     size = "medium",
@@ -142,19 +111,12 @@ py_test(
     deps = ["//:ray_lib"],
 )
 
-py_test(
-    name = "test_stress_sharded_direct",
-    size = "medium",
-    srcs = ["test_stress_sharded_direct.py", "test_stress_sharded.py"],
-    tags = ["exclusive"],
-    deps = ["//:ray_lib"],
-)
-
 py_test(
     name = "test_stress_failure",
     size = "large",
     srcs = ["test_stress_failure.py"],
-    tags = ["exclusive"],
+    # TODO(ekl) enable again once we support direct call reconstruction
+    tags = ["exclusive", "manual"],
     deps = ["//:ray_lib"],
 )
 
@@ -209,14 +171,6 @@ py_test(
     deps = ["//:ray_lib"],
 )
 
-py_test(
-    name = "test_failure_direct",
-    size = "medium",
-    srcs = ["test_failure_direct.py", "test_failure.py"],
-    tags = ["exclusive"],
-    deps = ["//:ray_lib"],
-)
-
 py_test(
     name = "test_garbage_collection",
     size = "small",
@@ -276,7 +230,8 @@ py_test(
     name = "test_monitors",
     size = "small",
     srcs = ["test_monitors.py"],
-    tags = ["exclusive"],
+    # TODO(ekl) tasks() and objects() are different in direct call mode.
+    tags = ["exclusive", "manual"],
     deps = ["//:ray_lib"],
 )
 
diff --git a/python/ray/tests/test_actor_direct.py b/python/ray/tests/test_actor_direct.py
deleted file mode 100644
index c302b4596..000000000
--- a/python/ray/tests/test_actor_direct.py
+++ /dev/null
@@ -1,16 +0,0 @@
-"""Wrapper script that sets RAY_FORCE_DIRECT."""
-
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
-
-import pytest
-import sys
-import os
-
-if __name__ == "__main__":
-    os.environ["RAY_FORCE_DIRECT"] = "1"
-    sys.exit(
-        pytest.main(
-            ["-v",
-             os.path.join(os.path.dirname(__file__), "test_actor.py")]))
diff --git a/python/ray/tests/test_actor_resources_direct.py b/python/ray/tests/test_actor_resources_direct.py
deleted file mode 100644
index 63263afc6..000000000
--- a/python/ray/tests/test_actor_resources_direct.py
+++ /dev/null
@@ -1,17 +0,0 @@
-"""Wrapper script that sets RAY_FORCE_DIRECT."""
-
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
-
-import pytest
-import sys
-import os
-
-if __name__ == "__main__":
-    os.environ["RAY_FORCE_DIRECT"] = "1"
-    sys.exit(
-        pytest.main([
-            "-v",
-            os.path.join(os.path.dirname(__file__), "test_actor_resources.py")
-        ]))
diff --git a/python/ray/tests/test_basic_direct.py b/python/ray/tests/test_basic_direct.py
deleted file mode 100644
index 028c46cfa..000000000
--- a/python/ray/tests/test_basic_direct.py
+++ /dev/null
@@ -1,16 +0,0 @@
-"""Wrapper script that sets RAY_FORCE_DIRECT."""
-
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
-
-import pytest
-import sys
-import os
-
-if __name__ == "__main__":
-    os.environ["RAY_FORCE_DIRECT"] = "1"
-    sys.exit(
-        pytest.main(
-            ["-v",
-             os.path.join(os.path.dirname(__file__), "test_basic.py")]))
diff --git a/python/ray/tests/test_failure_direct.py b/python/ray/tests/test_failure_direct.py
deleted file mode 100644
index 7e635d1df..000000000
--- a/python/ray/tests/test_failure_direct.py
+++ /dev/null
@@ -1,16 +0,0 @@
-"""Wrapper script that sets RAY_FORCE_DIRECT."""
-
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
-
-import pytest
-import sys
-import os
-
-if __name__ == "__main__":
-    os.environ["RAY_FORCE_DIRECT"] = "1"
-    sys.exit(
-        pytest.main(
-            ["-v",
-             os.path.join(os.path.dirname(__file__), "test_failure.py")]))
diff --git a/python/ray/tests/test_multinode_failures_direct.py b/python/ray/tests/test_multinode_failures_direct.py
deleted file mode 100644
index ac15f5d82..000000000
--- a/python/ray/tests/test_multinode_failures_direct.py
+++ /dev/null
@@ -1,18 +0,0 @@
-"""Wrapper script that sets RAY_FORCE_DIRECT."""
-
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
-
-import pytest
-import sys
-import os
-
-if __name__ == "__main__":
-    os.environ["RAY_FORCE_DIRECT"] = "1"
-    sys.exit(
-        pytest.main([
-            "-v",
-            os.path.join(
-                os.path.dirname(__file__), "test_multinode_failures.py")
-        ]))
diff --git a/python/ray/tests/test_stress_direct.py b/python/ray/tests/test_stress_direct.py
deleted file mode 100644
index 2ae68c439..000000000
--- a/python/ray/tests/test_stress_direct.py
+++ /dev/null
@@ -1,16 +0,0 @@
-"""Wrapper script that sets RAY_FORCE_DIRECT."""
-
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
-
-import pytest
-import sys
-import os
-
-if __name__ == "__main__":
-    os.environ["RAY_FORCE_DIRECT"] = "1"
-    sys.exit(
-        pytest.main(
-            ["-v",
-             os.path.join(os.path.dirname(__file__), "test_stress.py")]))
diff --git a/python/ray/tests/test_stress_sharded_direct.py b/python/ray/tests/test_stress_sharded_direct.py
deleted file mode 100644
index dab425e8f..000000000
--- a/python/ray/tests/test_stress_sharded_direct.py
+++ /dev/null
@@ -1,17 +0,0 @@
-"""Wrapper script that sets RAY_FORCE_DIRECT."""
-
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
-
-import pytest
-import sys
-import os
-
-if __name__ == "__main__":
-    os.environ["RAY_FORCE_DIRECT"] = "1"
-    sys.exit(
-        pytest.main([
-            "-v",
-            os.path.join(os.path.dirname(__file__), "test_stress_sharded.py")
-        ]))
diff --git a/python/ray/tune/trial_runner.py b/python/ray/tune/trial_runner.py
index cf749fb2e..1f2bc7ca4 100644
--- a/python/ray/tune/trial_runner.py
+++ b/python/ray/tune/trial_runner.py
@@ -14,6 +14,7 @@ import types
 
 import ray.cloudpickle as cloudpickle
 from ray.tune import TuneError
+from ray.tune.progress_reporter import trial_progress_str
 from ray.tune.ray_trial_executor import RayTrialExecutor
 from ray.tune.result import (TIME_THIS_ITER_S, RESULT_DUPLICATE,
                              SHOULD_CHECKPOINT)
@@ -380,7 +381,8 @@ class TrialRunner(object):
     def debug_string(self, delim="\n"):
         messages = [
             self._scheduler_alg.debug_string(),
-            self.trial_executor.debug_string()
+            self.trial_executor.debug_string(),
+            trial_progress_str(self.get_trials()),
         ]
         return delim.join(messages)
 
diff --git a/rllib/tests/test_eager_support.py b/rllib/tests/test_eager_support.py
index fd325aae1..a00e56add 100644
--- a/rllib/tests/test_eager_support.py
+++ b/rllib/tests/test_eager_support.py
@@ -73,4 +73,6 @@ class TestEagerSupport(unittest.TestCase):
 
 
 if __name__ == "__main__":
-    unittest.main(verbosity=2)
+    import pytest
+    import sys
+    sys.exit(pytest.main(["-v", __file__]))