ConnectionPool hangs in thin mode when timeout is not None

[lpetre-ulb]

Hello,

while porting my project to use a ConnectionPool instead of many Connection, I stumbled across an issue when:

• using the thin mode;
• setting the timeout to an integer (so not None);
• and using a dynamic number of connections (i.e. min != max).

In this specific case, the ConnectionPool.acquire() hangs and never returns a connection. I could identify the problem as due to to the following while loop. I'm not familiar enough with the library intrinsics to provide a PR though.

python-oracledb/src/oracledb/impl/thin/pool.pyx

Lines 329 to 333 in 7eb9be4

	while self.get_open_count() > self.min and conn_impls_to_check:
	conn_impl = conn_impls_to_check[0]
	if current_time - conn_impl._time_in_pool > self._timeout:
	conn_impls_to_check.pop(0)
	self._drop_conn_impl(conn_impl)

The project uses Python 3.8.13 with oracledb 1.3.1. The problem is however present since the feature implementation in c7c0d55.

Please find a minimal reproducible example below. The script blocks and never completes, staying in an endless loop in ConnectionPool.acquire().

import oracledb
from multiprocessing.pool import ThreadPool

pool = oracledb.create_pool("<REDACTED>", config_dir="/etc", min=1, max=5, timeout=10)
thread_pool = ThreadPool(processes = 5)

def wait(i):
  with pool.acquire() as connection:
    with connection.cursor() as cursor:
      cursor.execute("""
        BEGIN
          DBMS_SESSION.sleep(1);
        END; """)
      print(f"Done {i}")

for i in range(10):
  thread_pool.apply_async(wait, args=[i])

Thank you for any help. Let me know if you need more informations.

[anthony-tuininga]

Thanks for the report. This will get addressed in the next release!

[lpetre-ulb]

Wonderful! Thank you for your prompt answer.

[anthony-tuininga]

I have pushed a patch that should correct this issue. If you are able to build from source you can verify that it corrects your issue as well.

[anthony-tuininga]

This has been included in version 1.3.1 which was just released!

[lpetre-ulb]

Hello, all my apologies for the delayed reply. I confirm that the bug is indeed fixed in version 1.3.1. Thank you for the patch and the release.