if unset, set PYTHONHASHSEED in pre_spawn_env

dask · Dec 12, 2023 · 6f2b004 · 6f2b004
1 parent d249170
commit 6f2b004
Showing 1 changed file with 7 additions and 0 deletions.
diff --git a/distributed/nanny.py b/distributed/nanny.py
@@ -233,6 +233,13 @@ def __init__(  # type: ignore[no-untyped-def]
         self.Worker = Worker if worker_class is None else worker_class
 
         self.pre_spawn_env = _get_env_variables("distributed.nanny.pre-spawn-environ")
+        # To get consistent hashing on subprocesses, we need to set a consistent seed for
+        # the Python hash algorithm; xref https://github.com/dask/distributed/pull/8400
+        if self.pre_spawn_env.get("PYTHONHASHSEED") in (None, "0"):
+            # This number is arbitrary; it was chosen to commemorate
+            # https://github.com/dask/dask/issues/6640.
+            self.pre_spawn_env.update({"PYTHONHASHSEED": "6640"})
+
         self.env = merge(
             self.pre_spawn_env,
             _get_env_variables("distributed.nanny.environ"),