[Pgpool-general] Pgpool I vs Pgpool II

Toshihiro Kitagawa kitagawa at sraoss.co.jp
Thu Sep 22 09:57:19 UTC 2011


Hi Glyn, takizo,

I created a patch to fix a massive memory leak of raw mode.
The patch which was posted before by Yoshiyuki seems correct.
However, I think it's better to call pool_query_context_destroy()
than free_parser().

Would you try an attached patch?

The patch is for V3.0_STABLE branch, but it's applicable to
pgpool-II 3.0.x too.

Regards,

-- 
Toshihiro Kitagawa
SRA OSS, Inc. Japan

On Thu, 8 Sep 2011 20:14:41 +0900
Toshihiro Kitagawa <kitagawa at sraoss.co.jp> wrote:

> Hi Glyn, takizo,
> 
> Sorry for delay.
> 
> I fixed some memory leaks in pgpool-II 3.1 and V3_0_STABLE.
> But I guess memory leaks are remaining as you say.
> 
> Surely, free_parser() which was commented out might be problem.
> However, I think simply uncommenting leads unexpected new problem.
> So we have to fix those carefully.
> 
> I will investigate and fix memory problems which was posted to
> this ML from now on.
> 
> Regards,
> 
> -- 
> Toshihiro Kitagawa
> SRA OSS, Inc. Japan
> 
> On Thu, 8 Sep 2011 16:22:50 +0800
> takizo <paulooi at takizo.com> wrote:
> 
> > Glyn, 
> > 
> > Surely, I will try 2.3.3 :) 
> > Thanks a lot 
> > 
> > --
> > Paul Ooi 
> > 
> > 
> > 
> > On Sep 8, 2011, at 4:00 PM, Glyn Astill wrote:
> > 
> > > Hi takizo,
> > > 
> > > We are currently using Pgpool-II 2.3.3 and it has been really stable for us.  Unless you're after a specific feature in Pgpool-II 3.x you could perhaps try it?
> > > 
> > > So far I've had no luck with the patch posted by Yoshiyuki, even with that patch there is still a massive memory leak in 3.0.4.  It would appear that pool_proto_modules.c has lots commented out calls to free_parser(), many more than the single case that Yoshiyuki uncommented with his patch, so either there is a bit of a mess there that needs sorting out by someone who knows the overall logic, or the issue is elsewhere.
> > > 
> > > Glyn
> > > 
> > > From: takizo <paulooi at takizo.com>
> > > To: Glyn Astill <glynastill at yahoo.co.uk>
> > > Cc: "pgpool-general at pgfoundry.org" <pgpool-general at pgfoundry.org>
> > > Sent: Thursday, 8 September 2011, 1:48
> > > Subject: Re: [Pgpool-general] Pgpool I vs Pgpool II
> > > 
> > > Gyln, 
> > > 
> > > Thanks. I happened to read through the thread after posted the mail. It seems like I have to stick with Pgpool I until the developer is free. 
> > > 
> > > --
> > > takizo
> > > 
> > > 
> > > 
> > > On Sep 7, 2011, at 10:53 PM, Glyn Astill wrote:
> > > 
> > >> Hi takizo,
> > >> 
> > >> Yes this bug has been seen by numerous people on the list, Yoshiyuki Asaba has posted a patch, however there has been no response from the main Pgpool II developers.
> > >> 
> > >> I did ask a few times, and also asked if we'd be better testing 3.1, but alas still no response.  I can only assume the devs are busy or otherwise indisposed.
> > >> 
> > >> See the thread below for the patch.
> > >> 
> > >> http://www.mail-archive.com/pgpool-general@pgfoundry.org/msg03141.html
> > >> 
> > >> Glyn
> > >> 
> > >> From: takizo <paulooi at takizo.com>
> > >> To: pgpool-general at pgfoundry.org
> > >> Sent: Wednesday, 7 September 2011, 15:02
> > >> Subject: [Pgpool-general] Pgpool I vs Pgpool II
> > >> 
> > >> Hi all, 
> > >> 
> > >> Today I have upgraded Pgpool I to Pgpool II. And I had to roll back to Pgpool I due to memory hunger issue. 
> > >> 
> > >> My server is configured with 
> > >> PostgreSQL 8.2, 
> > >> running on FreeBSD 8.2
> > >> Intel Xeon box with 4 CPUs with Quad Core, total of 16 cores
> > >> Memory with 24GB
> > >> 
> > >> * While running on Pgpool I
> > >> I do no have memory hunger issue.  On postgresql.conf, I have max_connections=200 and shared_buffers=2GB. 
> > >> Most of the time, I have 8GB memory in used and 10GB Free and some in cached. Everything run just fine and perfectly good. 
> > >> 
> > >> * I tried upgraded to Pgpool II 
> > >> Same config on postgresql.conf, when I started PgPool II (3.0.4), my active memory started increasing and it didn't drop at all. 
> > >> I left the server running and after about 20 minutes, 20GB memory is in used, ended up it used swap memory because running out of memory. 
> > >> 
> > >> I run a reload/restart on pgpool, I gained back 14GB of free memory, and slowly it started to taken up those freed memory in active memory after about 15-20 minutes. 
> > >> I tried to bring down max_connections and shared_buffers value, and the memory doesn't stop taking up. It still slowly gaining all the memory it wants. 
> > >> 
> > >> Has anyone having that problem as well? I hope I can find some answer here :) 
> > >> Wanted to try out pgpool II performance but this problem is the stopper for now. 
> > >> 
> > >> Hope to get feedback from gurus in the house.
> > >> 
> > >> Thanks! 
> > >> 
> > >> --
> > >> takizo
> > >> _______________________________________________
> > >> Pgpool-general mailing list
> > >> Pgpool-general at pgfoundry.org
> > >> http://pgfoundry.org/mailman/listinfo/pgpool-general
> > >> 
> > >> 
> > > 
> > > 
> > > 
> > 
> 
> _______________________________________________
> Pgpool-general mailing list
> Pgpool-general at pgfoundry.org
> http://pgfoundry.org/mailman/listinfo/pgpool-general
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: memory_leak.patch
Type: application/octet-stream
Size: 9448 bytes
Desc: not available
URL: <http://pgfoundry.org/pipermail/pgpool-general/attachments/20110922/0649951b/attachment-0001.obj>


More information about the Pgpool-general mailing list