[Potential Bug] Unable to pass file to class constructor via `.remote`

Search before asking

• I searched the issues and found no similar issues.

Ray Component

Ray Core

What happened + What you expected to happen

I am trying to pass a file (or more specifically a file-like object on Windows) to a python class constructor via call to remote. I get a KeyError: file while doing so.

As a user I would expect to things out work fine because opening and reading/writing to a file is possible inside a class constructor but passing the file object of an already opened is not possible.

Is this prohibited due to possible race conditions?

Versions / Dependencies

Ray - master branch.

Python - 3.8.11

OS - Windows 10 Pro

Reproduction script

import ray, tempfile

exit_conditions = ["__ray_terminate__", "ray.actor.exit_actor", "ray.kill"]

@ray.remote
class A():
    def __init__(self, tmpfile, data):
        import atexit

        def f(*args, **kwargs):
            # with open(tmpfile, "w") as f:
            tmpfile.write(data)
            tmpfile.flush()

        atexit.register(f)

    def ready(self):
        pass

    def exit(self):
        ray.actor.exit_actor()

for exit_condition in exit_conditions:

    data = "hello"
    tmpfile = tempfile.NamedTemporaryFile("w+", suffix=".tmp")

    a = A.remote(tmpfile, data)
    ray.get(a.ready.remote())

    if exit_condition == "out_of_scope":
        del a
    elif exit_condition == "__ray_terminate__":
        ray.wait([a.__ray_terminate__.remote()])
    elif exit_condition == "ray.actor.exit_actor":
        ray.wait([a.exit.remote()])
    elif exit_condition == "ray.kill":
        ray.kill(a)
    else:
        assert False, "Unrecognized condition"

    def check_file_written():
        if tmpfile.read() == data:
            return True
        return False

    if exit_condition == "ray.kill":
        assert not check_file_written()

Error

2022-01-14 10:14:06,525 INFO services.py:1382 -- View the Ray dashboard at http://127.0.0.1:8265
 pid=7052) 2022-01-14 10:14:14,743      ERROR serialization.py:289 -- 'file'
 pid=7052) Traceback (most recent call last):
 pid=7052)   File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 287, in deserialize_objects
 pid=7052)     obj = self._deserialize_object(data, metadata, object_ref)
 pid=7052)   File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 194, in _deserialize_object
 pid=7052)     return self._deserialize_msgpack_data(data, metadata_fields)
 pid=7052)   File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 172, in _desTraceback (most recent call last):

  File "C:\Users\gagan\ray_project\file_pass.py", line 29, in <module>
 pid=7052)     python_objects = self._deserialize_pickle5_data(pickle5_data)
 pid=7052)   File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 162, in _deserialize_pickle5_data
    =7052)     obj = pickle.loads(in_band)
ray.get(a.ready.remote())
  File "c:\users\gagan\ray_project\ray\python\ray\_private\client_mode_hook.py", line 105, in wrapper
 pid=7052)   File "C:\ProgramData\Anaconda3\envs\ray_dev\lib\tempfile.py", line 467, in __getattr__
 pid=7052) func(*args, **kwargs)(A
  File "c:\users\gagan\ray_project\ray\python\ray\worker.py", line 1744, in get

    raise value
.pid=7052)ray.exceptions KeyError: 'file'
 pid=7052)The actor died because of an error raised in its creation task,  2022-01-14 10:14:14,759       ERROR worker.py:432 -- Exception raised in creation task: The actor died because of an error raised in its creation task, ray::A.__init__() (pid=7052, ip=127.0.0.1)
  File "python\ray\_raylet.pyx", line 596, in ray._raylet.execute_task
    with core_worker.profile_event(b"task:deserialize_arguments"):
  File "python\ray\_raylet.pyx", line 618, in ray._raylet.execute_task
    raise_if_dependency_failed(arg)
  File "python\ray\_raylet.pyx", line 450, in ray._raylet.raise_if_dependency_failed
    raise arg
ray.exceptions.RaySystemError: System error: 'file'
traceback: Traceback (most recent call last):
  File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 287, in deserialize_objects
    obj = self._deserialize_object(data, metadata, object_ref)
  File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 194, in _deserialize_object
    return self._deserialize_msgpack_data(data, metadata_fields)
  File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 172, in _deserialize_msgpack_data
    python_objects = self._deserialize_pickle5_data(pickle5_data)
  File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 162, in _deserialize_pickle5_data
    obj = pickle.loads(in_band)
  File "C:\ProgramData\Anaconda3\envs\ray_dev\lib\tempfile.py", line 467, in __getattr__
    file = self.__dict__['file']
KeyError: 'file'
ray::A.__init__() (pid=7052, ip=127.0.0.1)
 pid=7052)   File "python\ray\_raylet.pyx", line 596, in ray._raylet.execute_task
 pid=7052)     with core_worker.profile_event(b"task:deserialize_arguments"):
 pid=7052)   File "python\ray\_raylet.pyx", line 618, in ray._raylet.execute_task
 pid=7052)     raise_if_dependency_failed(arg)
 pid=7052)   File "python\ray\_raylet.pyx", line 450, in ray._raylet.raise_if_dependency_failed
 pid=7052)     raise arg
 pid=7052) ray.exceptions.RaySystemError: System error: 'file'
 pid=7052) traceback: Traceback (most recent call last):
 pid=7052)   File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 287, in deserialize_objects
 pid=7052)     obj = self._deserialize_object(data, metadata, object_ref)
 pid=7052)   File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 194, in _deserialize_object
 pid=7052)     return self._deserialize_msgpack_data(data, metadata_fields)
 pid=7052)   File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 172, in _deserialize_msgpack_data
 pid=7052)     python_objects = self._deserialize_pickle5_data(pickle5_data)
 pid=7052)   File "c:\users\gagan\ray_project\ray\python\ray\serialization.py", line 162, in _deserialize_pickle5_data
 pid=7052)     obj = pickle.loads(in_band)
 pid=7052)   File "C:\ProgramData\Anaconda3\envs\ray_dev\lib\tempfile.py", line 467, in __getattr__
 pid=7052)     file = self.__dict__['file']
 pid=7052) KeyError: 'file'

Anything else

No response

Are you willing to submit a PR?

• Yes I am willing to submit a PR!

[pcmoritz]

I think it would make to give a better error message here, but it is fine not to support it, since this won’t work on a cluster with many machines for example.

[czgdp1807]

why is the tmpfile getting into the serialized data?

A use case I discovered while fixing https://github.com/ray-project/ray/pull/21484

If passing file-like or file objects is prohibited then we should capture it earlier and raise a TypeError/ValueError for easier understanding.