Is it odd that my SQLAlchemy MySQL connection always ends up sleeping?


I have a Flask web application using SQLAlchemy with MySQL, and I have set up a scopedsession(). I also have a teardown_request handler that calls session.remove() after every request is finished. For some odd reason, if no requests are made on the web app for a day or more, the app gets "Operationalerror: MySQL Server has gone away".

In my debugging mission, I looked at SHOW PROCESSLIST and saw the following:

39817253 | sqladmin | my_host | kb_dev   | Sleep   |  174 |

The 174 is the number of seconds the connection from my application has been "sleeping". It keeps counting up if the application doesn't make another request.

It seems like my application holds on to the connection to MySQL even after my request has finished! And there is usually only one process no matter how many requests I make with my application, simultaneously or not.

My question is if it is normal for the connection to be "sleeping" this long? I'm pretty sure the extended sleeping is causing MySQL to cut the connection after a certain timeout which in turn is causing the "OperationalError: Mysql has gone away" error.

1/11/2012 6:42:50 PM

Accepted Answer

SQLAlchemy's default behavior is to pool connections within the Engine:

As far as the disconnect overnight thing, this is a known MySQL behavior, SQLAlchemy provides the pool_recycle flag to work around it. Here are many links which describe it: (pool_recycle)

Blog post from just a few days ago:

1/11/2012 6:56:19 PM

Licensed under: CC-BY-SA with attribution
Not affiliated with: Stack Overflow