|
From: Scott R. <sc...@to...> - 2004-03-02 23:55:26
|
Presto, seems to hold up just fine now. Performance appears improved,
as well.
Once again, for anyone searching the mailinglist, in DBConnection.py,
def iterSelect(self, select):
conn = self.getConnection()
return self._iterSelect(conn, select, self, False)
On Tue, 2004-03-02 at 18:17, Ian Bicking wrote:
> Scott Russell wrote:
> > David and Ian:
> >
> > Yes, (as you'll see in my mail that arrived the same time yours did) it
> > looks like a thread concurrency issue.
> >
> > If I turn on debugging, the threading test survives many additional
> > threads, but will die eventually. (David, 32 kills it very handily.)
> > Whether because of some serialization that gets forced during printing,
> > because it slows the process down enough to avoid contention, or some
> > third issue, I can't say.
> >
> > Here is a sample run, with 32 threads. Sometimes it lasts for 30-40
> > queries, sometimes (as seen here) only 3....
> >
> > $ python Main/TestPage.py
> > 1:Dummy-1/Select : SELECT person.id, person.email, person.firstname, person.lastname FROM person WHERE (person.email = 'xxx') LIMIT 1
> > 1:Dummy-2/Select : SELECT person.id, person.email, person.firstname, person.lastname FROM person WHERE (person.email = 'xxx') LIMIT 1
> > 1:Dummy-3/Select : SELECT person.id, person.email, person.firstname, person.lastname FROM person WHERE (person.email = 'xxx') LIMIT 1
> > Killed
> >
> > What next?
>
> Well, it's reusing the same connection each time. Which it probably
> shouldn't be -- it can reuse them sometimes, but you'd think it would
> need to do otherwise sometimes.
>
> Hmm...
>
> Bah, DBConnection.iterSelect is dumb. It shouldn't be written like it
> is. Right now it looks like:
>
> def iterSelect(self, select):
> return self._runWithConnection(self._iterSelect, select, self,
> False)
>
> Which means the connection is returned to the pool immediately when the
> generator is returned, then reused when it shouldn't be, and worse it
> can be returned to the pool another time if the select results are
> exhausted. It should look like:
>
> def iterSelect(self, select):
> conn = self.getConnection()
> return self._iterSelect(conn, select, self, False)
>
> This isn't quite right either, as the connection will only be returned
> if you exhaust the select. _iterSelect should be broken out into an
> object that returns the connection on __del__. Anyway, this iterSelect
> should resolve your concurrency problems for the moment (I haven't
> tested it, though; report back if it works)
>
> Ian
>
>
> -------------------------------------------------------
> SF.Net is sponsored by: Speed Start Your Linux Apps Now.
> Build and deploy apps & Web services for Linux with
> a free DVD software kit from IBM. Click Now!
> http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
> _______________________________________________
> sqlobject-discuss mailing list
> sql...@li...
> https://lists.sourceforge.net/lists/listinfo/sqlobject-discuss
|
