fix(aiomysql): avoid wrapping pooled connections multiple times

[canonrock16]

Overview

• prevent repeatedly wrapping aiomysql pooled connections with AsyncConnectionWrapper
• avoid building deep ObjectProxy chains that led to RecursionError: maximum recursion depth exceeded
• add a regression unit test (tests/agent_unittests/test_database_aiomysql.py) that simulates a pool returning the same connection twice

Related Github Issue

None.
Related forum thread: https://support.newrelic.com/s/hubtopic/aAXPh000000EHIjOAO/recursionerror-when-new-relic-python-agent-instruments-aiomysql-still-happening

Testing

• tox -c tox.ini -e python-agent_unittests-py312-without_extensions
  All tests passed including the new test I added

[CLAassistant]

All committers have signed the CLA.

[TimPansino]

I'll handle moving the tests for this to the right place and making sure they're working, thanks for finding this though.

Really curious how Pool._acquire() is maintaining the wrappers such that they're preserved for future calls. I'll have to read the codebase again and look into it.

[canonrock16]

Reproduction details

Agent 10.17.0 / aiomysql 0.2.0 / Python 3.12
Applied load against the service with hey; during the run the service began returning 500s and CloudWatch logs filled with RecursionError: maximum recursion depth exceeded
Full write-up: https://support.newrelic.com/s/hubtopic/aAXPh000000EHIjOAO/recursionerror-when-new-relic-python-agent-instruments-aiomysql-still-happening

Sample log :

2025-10-22T08:52:10.099Z RecursionError: maximum recursion depth exceeded
2025-10-22T08:52:10.184Z ERROR:root:unexpected error
File "<internal_path>/server.py", line 105, in wrap_error
return await func(*args, **kwargs)
...
File "<internal_path>/newrelic/hooks/database_aiomysql.py", line 79, in _wrap_pool__acquire
connection = await wrapped(*args, **kwargs)
File "<internal_path>/aiomysql/pool.py", line 171, in _fill_free_pool
self._loop.time() - conn.last_usage > self._recycle)
File "<internal_path>/newrelic/common/object_wrapper.py", line 70, in getattr
return super().getattr(name)
[previous line repeated 322 more times]
RecursionError: maximum recursion depth exceeded

After this patch, the same load no longer produces RecursionErrors and the service stays healthy.

[canonrock16]

I'll handle moving the tests for this to the right place and making sure they're working, thanks for finding this though.

Really curious how Pool._acquire() is maintaining the wrappers such that they're preserved for future calls. I'll have to read the codebase again and look into it.

For reference, aiomysql’s pool implementation reuses the same connection object pulled from the free queue:

• _acquire: https://github.com/aio-libs/aiomysql/blob/v0.2.0/aiomysql/pool.py#L118-L138
• _fill_free_pool: https://github.com/aio-libs/aiomysql/blob/v0.2.0/aiomysql/pool.py#L140-L196

_acquire() pops a connection from self._free, and _fill_free_pool() rotates existing entries unless they’re closed or expired. As a result, the same instance that was returned to the pool is handed back on the next acquire(), so the agent’s wrapper stays attached unless we detect it and avoid wrapping again.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 81.76%. Comparing base (fcaeb10) to head (5887526).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1553      +/-   ##
==========================================
- Coverage   81.76%   81.76%   -0.01%     
==========================================
  Files         207      207              
  Lines       23946    23948       +2     
  Branches     3796     3797       +1     
==========================================
+ Hits        19580    19581       +1     
  Misses       3100     3100              
- Partials     1266     1267       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.