Re: Multi-threaded Remote Calls vs. Python imterpreter Lock

On 16/03/2008, Jozsef Szalay <jszalay-7JS02rSrduhWk0Htik3J/[email protected]> wrote:
> Please ignore my previous email! Apparently I was wrong. (Not the first
>  time :-)) My third thread eventually got blocked regardless of what
>  cursor factory was used. Also, the cursors were serialized as James
>  suggested.

They need to be synchronised if they are associated with the same
transaction.  Imagine executing the following two statements in the
cursors:
    INSERT INTO sometable VALUES (...)
    SELECT * from sometable

The result of the SELECT query is going to depend on the first query:
you can't execute them in parallel.

If you really want to do work in parallel, you'll need multiple
connections which also means multiple transactions.


>  So the issue again is why the python interpreter lock is not released
>  when a second thread attempts a psycopg call on a connection that is
>  already busy?

Did you try any of the things from my previous email?  I'll try and
look into it when I have time, but it would be helpful if you could
provide a simple test program that exhibits the problem.

James.