Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat(python): Raise an error when users try to use Polars API in a fork()-without-execve() child #19149

Merged
merged 4 commits into from
Oct 10, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 35 additions & 0 deletions py-polars/polars/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -414,3 +414,38 @@ def __getattr__(name: str) -> Any:

msg = f"module {__name__!r} has no attribute {name!r}"
raise AttributeError(msg)


# fork() breaks Polars thread pool. Instead of silently hanging when users do
# this, e.g. by using multiprocessing's footgun default setting on Linux, warn
# them instead:
def __install_postfork_hook() -> None:
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nice fix. Any idea of the startup cost of this?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I suspect it's not meaningfully slower, but I'll measure it tomorrow.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This appears to add 3 or 4 µs to import polars. This doesn't seem like a problem to me.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Alright. Thanks @itamarst. Nice that we can catch those deadlocks now.

def fail(*args: Any, **kwargs: Any) -> None:
message = """\
Using fork() can cause Polars to deadlock in the child process.
In addition, using fork() with Python in general is a recipe for mysterious
deadlocks and crashes.

The most likely reason you are seeing this error is because you are using the
multiprocessing module on Linux, which uses fork() by default. This will be
fixed in Python 3.14. Until then, you want to use the "spawn" context instead.

See https://docs.pola.rs/user-guide/misc/multiprocessing/ for details.
"""
raise RuntimeError(message)

def post_hook_child() -> None:
# Switch most public Polars API to fail when called. This won't catch
# _all_ edge cases, but does make it more likely users get told they
# tried to do something broken.
for name in __all__:
if callable(globals()[name]):
globals()[name] = fail

import os

if hasattr(os, "register_at_fork"):
os.register_at_fork(after_in_child=post_hook_child)


__install_postfork_hook()
21 changes: 21 additions & 0 deletions py-polars/tests/unit/test_polars_import.py
Original file line number Diff line number Diff line change
@@ -1,6 +1,8 @@
from __future__ import annotations

import compileall
import multiprocessing
import os
import subprocess
import sys
from pathlib import Path
Expand Down Expand Up @@ -97,3 +99,22 @@ def test_polars_import() -> None:
import_time_ms = polars_import_time // 1_000
msg = f"Possible import speed regression; took {import_time_ms}ms\n{df_import}"
raise AssertionError(msg)


def run_in_child() -> pl.Series:
return pl.Series([1, 2, 3])


@pytest.mark.skipif(not hasattr(os, "fork"), reason="Requires fork()")
def test_fork_safety() -> None:
# Using fork()-based multiprocessing shouldn't work:
with (
multiprocessing.get_context("fork").Pool(1) as pool,
pytest.raises(RuntimeError, match=r"Using fork\(\) can cause Polars"),
):
pool.apply(run_in_child)

# Using forkserver and spawn context should not error out:
for context in ["spawn", "forkserver"]:
with multiprocessing.get_context(context).Pool(1) as pool:
pool.apply(run_in_child)
Loading