Operational Error When Silk Is Used On Big SQL Queries

[hanleyhansen]

What

Keep trying the _process_response until it works. It tends to fail with big MySQL queries.

Why

MySQLdb specifically does not manage the connections and will error if you try to reuse a stale connection. One that has closed or expired. You need to explicitly check for this error case and handle it. This PR excepts the OperationalError whenever the silk middleware has to processs a relatively large query in it's _process_response.

[hanleyhansen]

@avelis we can discuss on where I put the try catch and maybe think about moving it somewhere closer to the actual cause inside of so _process_response we don't mask any other errors. Let me know your thoughts.

[avelis]

@hanleyhansen Could there ever be a scenario where you could get forever stuck in the while loop? I will likely merge but wanted to poke any holes at it before I did.

In terms of placement. I think it's fine.

[hanleyhansen]

@avelis straight from the MySQL docs

This exception is raised for errors which are related to MySQL's operations. For example: too many connections; a host name could not be resolved; bad handshake; server is shutting down, communication errors.

So the only way we can get stuck in a loop forever is if the DB server goes down. At the point the person has a different problem though. Too many connections will keep trying until there are some available connections. Communication errors, same thing. Other stuff in Django will break well before silk gets stucked in a loop forever. Otherwise just catching OperationalError and AttributeError should be safe within the while True: loop.

[avelis]

@hanleyhansen Thanks for the contribution and providing a solution for this scenario!