Bulk load connection timeouts on 1.7.1

[amachanic]

Describe the bug

As soon as I updated to 1.7.1 (from 1.6.x) I started experiencing occasional errors when attempting to bulk load: "RuntimeErorr: Failed to connect to SQL Server: Timeout Error: Timeout while connecting"

I'm not really sure how to create a repro of the issue as it happened seemingly randomly. What I can say is that the code in question makes heavy use of multiprocessing, often running concurrent bulk loads on several processes at once. This setup has been rock solid on 1.6.x and I didn't change any other dependencies, so I'm very confident that this is a regression.

Further technical details

Python version: 3.12
SQL Server version: SQL Server 2022
Operating system: Windows 11

[github-actions]

Hi @amachanic, thank you for opening this issue!

Our team will review it shortly. We aim to triage all new issues within 24-48 hours and get back to you.

If you have additional information to share, please feel free to update the issue.

Thank you for your patience!

[amachanic]

Update: I tested some more using the code below. Running the code on Windows 11 with mssql-python 1.7.1, I am able to consistently see a failure. (I had to bump around some of the numbers a bit, so this may depend on instance size.) If I run the same either with mssql-python 1.6.0 or either version on Linux (WSL, Ubuntu LTS), I have yet to see a failure.

So, bottom line, this issue appears to be isolated to Windows, and it is definitely a mssql-python 1.7.x issue.

import multiprocessing
import mssql_python


_CONN_STR = [put your connection string here]


def t():
    conn = mssql_python.connect(_CONN_STR, autocommit=True, timeout=10)
    cur = conn.cursor()

    cur.bulkcopy(
        "##test", ((r, r, r) for r in range(100000)),
        timeout=1800
    )


if __name__ == '__main__':
    conn = mssql_python.connect(_CONN_STR, autocommit=True, timeout=10)
    cur = conn.cursor()

    cur.execute("DROP TABLE IF EXISTS ##test; CREATE TABLE ##test(i int, j int, k int);")

    with multiprocessing.Pool(processes=32) as pool:
        pool.starmap(t, [() for _ in range(128)])

    cur.execute("SELECT COUNT(*) FROM ##test")
    print(cur.fetchall())

[amachanic]

And followup question that just occurred to me: What exactly does bulkcopy use for its connection timeout? Is it inherited from the outer connection object? (Which would be my assumption.) Or is there some other number in there somewhere? I don't see any obvious property to directly manipulate it.