Trac connection pool depletion when running multithreaded

[alberto@…]

I have a trac mounted via WSGI in a multithreaded CherryPy HTTP server and after a few dozen requests the db connections pool seems to get exhausted and Trac freezes up.

Other WSGI apps mounted on the same CP server suffer no lockup while Trac remains locked-up.

I've managed to reproduce both using sqlite and postgres backends.

This message from Trac gets on my server's log:

Exception exceptions.AssertionError: <exceptions.AssertionError instance at 0xa931db4c> in <bound method PooledConnection.__del__ of <trac.db.pool.PooledConnection object at 0xa936b76c>> ignored

Which I believe to be located ​here

I'm not too familiar with Trac's pool management to submit a patch ATM, sorry.

[Matthew Good]

This is a duplicate of #3504 which should be fixed in the trunk now.

[Christian Boos]

Please alberto, can you try to reproduce the issue with latest trunk, to be really sure that it's fixed now?

[alberto]

Upgraded to trunk and it survived 1000 requests in chunks of 50 concurrent requests so it surely seems fixed. Thanks! :)

Alberto

[Christian Boos]

Great! Thank you very much for the feedback.