Re: Join query including two generate_series calls causes big memory growth and crash

On Wed, Apr 20, 2011 at 5:34 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> =?ISO-8859-1?Q?Jorge_Ar=E9valo?= <jorge(dot)arevalo(at)deimos-space(dot)com> writes:
>> I'm executing this query:
>
>> SELECT x, y, another_field FROM generate_series(1, 10) x,
>> generate_series(1, 10) y, my_table
>
>> The field 'another_field' belongs to 'my_table'. And that table has
>> 360000 entries. In a 64 bits machine, with 4GB RAM, Ubuntu 10.10 and
>> postgres 8.4.7, the query works fine. But in a 32 bits machine, with
>> 1GB RAM, Ubuntu 9.10 and postgres 8.4.7, the query process is killed
>> after taking about 80% of available memory. In the 64 bits machine the
>> query takes about 60-70% of the available memory too, but it ends.
>
> You mean the backend, or psql?  I don't see any particular backend bloat
> when I do that, but psql eats memory because it's trying to absorb and
> display the whole query result.
>

Yes, the memory eater is psql, not backend.

>> Is it normal? I mean, postgres has to deal with millions of rows, ok,
>> but shouldn't it start swapping memory instead of crashing? Is a
>> question of postgres configuration?
>
> Try "\set FETCH_COUNT 1000" or so.
>
>                        regards, tom lane
>

Thanks for the tip.

Best regards,

--
Jorge Arévalo
Internet & Mobilty Division, DEIMOS
jorge(dot)arevalo(at)deimos-space(dot)com
http://es.linkedin.com/in/jorgearevalo80
http://mobility.grupodeimos.com/
http://gis4free.wordpress.com
http://geohash.org/ezjqgrgzz0g