Re: Running out of memory while making a join

Hi Tom,

Thank you for the analyzes!

No problem, there is no problem to use "select wm_nfsp.*" but as my concern
is to prevent this in the future I think I should apply the fix or is there
a config parameter to abend the backend if it reaches some kind of storage
limit?

Thank you!

Reimer

On Tue, Nov 13, 2012 at 5:51 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> Carlos Henrique Reimer <carlos(dot)reimer(at)opendb(dot)com(dot)br> writes:
> > That is what I got from gdb:
>
> > ExecutorState: 11586756656 total in 1391 blocks; 4938408 free (6
> > chunks); 11581818248 used
>
> So, query-lifespan memory leak. After poking at this for a bit, I think
> the problem has nothing to do with joins; more likely it's because you
> are returning a composite column:
>
> select wm_nfsp from "5611_isarq".wm_nfsp ...
>
> I found out that record_out() leaks sizable amounts of memory, which
> won't be recovered till end of query. You could work around that by
> returning "select wm_nfsp.*" instead, but if you really want the result
> in composite-column notation, I'd suggest applying this patch:
>
> http://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=c027d84c81d5e07e58cd25ea38805d6f1ae4dfcd
>
> regards, tom lane
>

--
Reimer
47-3347-1724 47-9183-0547 msn: carlos(dot)reimer(at)opendb(dot)com(dot)br