Re: Backend memory leakage when inserting

On Tue, 6 May 2003, Gerhard Hintermayer wrote:

> Gerhard Hintermayer wrote:
> > Should have investigated this better before posting:
> >
> > I have the following test case:
> > PostgreSQL 7.3.1 on i686-pc-linux-gnu, compiled by GCC 2.95.2 (running
> > a 2.2.23 kernel)
> > create an empty database
> > create table testdb(i integer,t timestamp);
> >
> > edit a file fill.sql with 500 identical lines:
> > INSERT INTO testdb VALUES(1,'now()');
> >
> > check memory of backend process (ps axuw): Column RSS 3224
> > \i fill.sql
> > backend RSS 3295 (ok, some sort of basic mem probably)
> > \i fill.sql ~ 10 times
> > backend RSS 3300
> > \i fill.sql ~ 25 times
> > backend RSS 3304
> > \i fill.sql ~ 70 times
> > backend RSS 3316
> >
> > VSZ value is always the same (6248). So what's going on here ?
> >
> >
> > (now i have ~ 53500 tuples in my table)
> >
> >
> > I have no idea, what's going on here ? Any ideas ?
> >
> > Gerhard
>
> Did compile postgreSQL (7.3.1) on a SuSE 8.1 box (Linux,2.4.19 kernel)
> and see exactly the same behavior. Seems like the amount of leakage
> depends on the size (number of tables, number of triggers ???) of the
> whole database, on a empty database the memory growth rate is quite slow
> (see posted example) , on the production system the mem usage was ~ 120
> MB after ~ 300000 inserts.

How many shared buffers is your machine set to use? it may just be using
them and it looks like a memory leak when it isn't.