Kaggle/docker-python

Interrupt code breaks sometimes

Open

#1338 opened on Dec 8, 2023

View on GitHub
 (1 comment) (0 reactions) (0 assignees)Python (2,260 stars) (946 forks)batch import
bughelp wanted

Description

🐛 Bug

When running certain code it becomes impossible to interrupt the kernel until that code completes.

To Reproduce

import sklearn
# get large dataset on the order of 1e6 or 1e7
df = ...
sklearn.preprocessing.QuantileTransformer(subsample=1e7, n_quantiles=1e6).fit_transform(df)

Expected behavior

It should interrupt!

Additional context

That's just an example, I've seen it plenty of times over the last few weeks, usually for commands I'd expect to take a while but that I want to stop midway to tweak something. That just seemed like an easy to replicate example. Oddly enough, the QuantileTransformer will interrupt just fine when n_quantiles is an order of magnitude smaller.

Contributor guide

Tech stack
pythondocker
Domain
infrastructureperformance
Issue type
bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
4
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
3-5 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
stale
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
clear
Prerequisites
PythonDockersignal handling
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
25
Research direction
Reproduce the bug inside the Kaggle Python Docker container using the provided code with large datasets. Investigate signal handling in Python and how the Docker container propagates signals to processes. Look at the Dockerfile and entrypoint scripts for any signal related configurations. Check if sklearn's QuantileTransformer runs C extensions that block the GIL, preventing interrupts. Consider if using Python's signal.setitimer or threading can mitigate the issue. Review similar issues in the repository or upstream.