Skip to content

[SPARK-54456][PYTHON] Import worker module after fork to avoid deadlock #53166

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.

Sign up for GitHub

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

Closed
wants to merge 1 commit into from
Closed

[SPARK-54456][PYTHON] Import worker module after fork to avoid deadlock #53166

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
21 changes: 13 additions & 8 deletions python/pyspark/daemon.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,14 +31,6 @@

from pyspark.serializers import read_int, write_int, write_with_length, UTF8Deserializer

if len(sys.argv) > 1 and sys.argv[1].startswith("pyspark"):
import importlib

worker_module = importlib.import_module(sys.argv[1])
worker_main = worker_module.main
else:
from pyspark.worker import main as worker_main


def compute_real_exit_code(exit_code):
# SystemExit's code can be integer or string, but os._exit only accepts integers
Expand Down Expand Up @@ -78,6 +70,19 @@ def worker(sock, authenticated):
return 1

exit_code = 0

# We don't know what could happen when we import the worker module. We have to
# guarantee that no thread is spawned before we fork, so we have to import the
# worker module after fork. For example, both pandas and pyarrow starts some
# threads when they are imported.
if len(sys.argv) > 1 and sys.argv[1].startswith("pyspark"):
import importlib

worker_module = importlib.import_module(sys.argv[1])
Copy link
Member

@ueshin ueshin Nov 24, 2025

Choose a reason for hiding this comment

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

Just a QQ: Is it ok to call this multiple times? This worker function will be called every UDF execution.

Copy link
Contributor Author

@gaogaotiantian gaogaotiantian Nov 24, 2025

Choose a reason for hiding this comment

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

Yes, the imported module will be cached, just like the normal import keyword.

worker_main = worker_module.main
else:
from pyspark.worker import main as worker_main

try:
worker_main(infile, outfile)
except SystemExit as exc:
Expand Down