[tune/rllib] Ignore directory exists errors to tackle race conditions

python/ray/tune/impl/tuner_internal.py

@@ -258,7 +258,7 @@ def _setup_create_experiment_checkpoint_dir(
             run_config.name,
         )
         if not os.path.exists(path):
-            os.makedirs(path)
+            os.makedirs(path, exist_ok=True)
         return path
 
     # This has to be done through a function signature (@property won't do).

rllib/algorithms/algorithm.py

@@ -260,7 +260,9 @@ class directly. Note that this arg can also be specified via
             timestr = datetime.today().strftime("%Y-%m-%d_%H-%M-%S")
             logdir_prefix = "{}_{}_{}".format(str(self), env_descr, timestr)
             if not os.path.exists(DEFAULT_RESULTS_DIR):
-                os.makedirs(DEFAULT_RESULTS_DIR)
+                # Possible race condition if dir is created several times on
+                # rollout workers
+                os.makedirs(DEFAULT_RESULTS_DIR, exist_ok=True)
             logdir = tempfile.mkdtemp(prefix=logdir_prefix, dir=DEFAULT_RESULTS_DIR)
 
             # Allow users to more precisely configure the created logger

rllib/offline/json_writer.py

@@ -60,10 +60,7 @@ def __init__(
         else:
             path = os.path.abspath(os.path.expanduser(path))
             # Try to create local dirs if they don't exist
-            try:
-                os.makedirs(path)
-            except OSError:
-                pass  # already exists
+            os.makedirs(path, exist_ok=True)
             assert os.path.exists(path), "Failed to create {}".format(path)
             self.path_is_uri = False
         self.path = path

rllib/policy/tf_policy.py

@@ -1,4 +1,3 @@
-import errno
 import logging
 import math
 import os
@@ -529,12 +528,7 @@ def export_checkpoint(
         self, export_dir: str, filename_prefix: str = "model"
     ) -> None:
         """Export tensorflow checkpoint to export_dir."""
-        try:
-            os.makedirs(export_dir)
-        except OSError as e:
-            # ignore error if export dir already exists
-            if e.errno != errno.EEXIST:
-                raise
+        os.makedirs(export_dir, exist_ok=True)
         save_path = os.path.join(export_dir, filename_prefix)
         with self.get_session().graph.as_default():
             saver = tf1.train.Saver()

rllib/policy/torch_policy.py

@@ -876,7 +876,7 @@ def export_model(self, export_dir: str, onnx: Optional[int] = None) -> None:
         }
 
         if not os.path.exists(export_dir):
-            os.makedirs(export_dir)
+            os.makedirs(export_dir, exist_ok=True)
 
         seq_lens = self._dummy_batch[SampleBatch.SEQ_LENS]
         if onnx:

rllib/policy/torch_policy_v2.py

@@ -940,7 +940,7 @@ def export_model(self, export_dir: str, onnx: Optional[int] = None) -> None:
         }
 
         if not os.path.exists(export_dir):
-            os.makedirs(export_dir)
+            os.makedirs(export_dir, exist_ok=True)
 
         seq_lens = self._dummy_batch[SampleBatch.SEQ_LENS]
         if onnx: