Re: possible GDI resource leak

From: "Magnus Hagander" <mha(at)sollentuna(dot)net>
To: "Gary Doades" <gpd(at)gpdnet(dot)co(dot)uk>, <pgsql-hackers-win32(at)postgresql(dot)org>
Subject: Re: possible GDI resource leak
Date: 2004-06-23 09:24:04
Message-ID: 6BCB9D8A16AC4241919521715F4D8BCE34BD39@algol.sollentuna.se
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers-win32

> > I'm running a recent snapshot and started noticing strange
> behavior on
> > the desktop. Explorer was refusing to hide/show windows and
> > right-clicking on the application icon in the explorer bar
> refused to
> > show the system menu. For those of you not in the know,
> this is the
> > classic sign that your system is out of handles (yes, this can even
> > happen on 2000/XP).
> >
> > I began killing processes one by one to see who it was and
> everything
> > returned to normal after shutting down the server. This could have
> > been a fluke, but be sure to mention it if you experience
> something similar.
> >
> > Merlin
>
> Why would postgres use GDI at all (apart from maybe a hidden
> console window and system menu)? Using Task Mangler I can see
> that each postgres instance uses 4 GDI objects. So unless you
> have a hell of a lot of connections to the database I can't
> see it using up a lot of GDI objects. OTOH each postgres
> instance uses 140 (ish) handles. So after a while handles may
> get a little clogged up if you have loads of connections.

Run postgresql for a while and it will use a lot more than 140 handles
(if you have a large database). IIRC, postgresql will set the limit at
1024 open files / process. With 100 backends, you're up to 100k
handles...

But yeah, pg should not use any GDI objects. And I've also noticed the
problem with the mingw console not closing correctly etc, so it could
very well come from there.

Another option is that it's the total handle pool that's dying. If it
happens again, plz check out what kind of handles you're high on (you
can use performance monitor to get it more granular, IIRC).

//Magnus

Browse pgsql-hackers-win32 by date

  From Date Subject
Next Message Merlin Moncure 2004-06-23 12:04:59 Re: backslash commands under win32
Previous Message Dave Page 2004-06-23 07:53:33 Re: [pgsql-hackers-win32] initdb regression ?