(forw) RE: [GENERAL] Memory leak in FreeBSD? - Mailing list pgsql-general

This doesn't look like a memory leak to me, can anyone working on
backend/libpq comment?

-Alfred

----- Forwarded message from Jeff Eckermann <jeckermann@verio.net> -----

Message-ID: <79103FDEF940D2118D9D00A0C9C9309CA35FA0@NEZU>
From: Jeff Eckermann <jeckermann@verio.net>
To: "'Alfred Perlstein'" <bright@wintelcom.net>
Subject: RE: [GENERAL] Memory leak in FreeBSD?
Date: Tue, 11 Jan 2000 16:30:43 -0600
X-Mailer: Internet Mail Service (5.5.2650.21)

Can we take this step by step?  I just tried the same query again, as
follows:

select distinct username, date (connect) as connect_date, date (disconnect)
as disconnect_date, timespan_to_seconds (elapsed) as seconds, inoct, outoct,
calling, called into dec_dist from december;
pqReadData() -- backend closed the channel unexpectedly.
        This probably means the backend terminated abnormally
        before or while processing the request.
We have lost the connection to the backend, so further processing is
impossible.  Terminating.
~
~
~
jeffe=> select count (*) from december;
  count
-------
2497190
(1 row)
I am estimating that the 'distinct' will select perhaps one quarter, or
less, of the total records.

> -----Original Message-----
> From:    Alfred Perlstein [SMTP:bright@wintelcom.net]
> Sent:    Tuesday, January 11, 2000 2:24 PM
> To:    Jeff Eckermann
> Cc:    'admin'; 'pgsql-general@postgresql.org'
> Subject:    Re: [GENERAL] Memory leak in FreeBSD?
>
> * Jeff Eckermann <jeckermann@verio.net> [000111 11:37] wrote:
> >
> > > -----Original Message-----
> > > From:    admin [SMTP:admin@wtbwts.com]
> > > Sent:    Tuesday, January 11, 2000 12:50 PM
> > > To:    Jeff Eckermann
> > > Cc:    'pgsql-general@postgresql.org'
> > > Subject:    RE: [GENERAL] Memory leak in FreeBSD?
> > >
> > > What is maxusers set to in your kernel? One prolem I had was that
> > > postgresql was using more filedescriptors that my kernel could handle.
> If
> > > you'd like to check your current filedescriptor status and your max,
> try:
> > > pstat -T. If that is your problem, change your maxusers to a suitable
> > > number and recompile your kernel.
> > >
> >
> > Maxusers is set to 128.  RAM is 256Mg.
> > Do you think this could be the problem?
>
> Saying it's a memory leak without describing any other sort of symptoms is
> not a very useful bug report.  Twiddling maxusers should have _no_ effect
> on whether an application leaks memory or not.
>
> So how about you explain the symptoms of the 'leak' (kernel messages,
> top, systat -vmstat) and what exactly you mean by it.  Does the postmaster
> gradually increase in memory size until the machine starts swapping?  If
> not, then it's probably _not_ a memory leak.
>
> thanks,
> -Alfred Perlstein - [bright@rush.net|alfred@freebsd.org]
> Wintelcom systems administrator and programmer
>    - http://www.wintelcom.net/ [bright@wintelcom.net]
>
> > > > FreeBSD port: I don't know enough to know what difference that might
> > > make.
> > > > Any suggestion you have would be appreciated: thanks.
> > > >
> > > > > Did you upgrade from source or from the freebsd ports?
> > > > >
> > > > > > We upgraded to version 6.5.2 recently, running on FreeBSD 3.0.
> Now
> > > we
> > > > > are
> > > > > > having problems with moderately complex queries failing to
> complete
> > > > > (backend
> > > > > > terminating unexpectedly; last one crashed the server).  The
> most
> > > likely
> > > > > > explanation appears to be a memory leak.  Is there any known
> problem
> > > > > with
> > > > > > FreeBSD?
> > > >
> > >
> > >
> > > ************
> >
> > ************

----- End forwarded message -----