Re: Large join runs out of memory in 8.1

On Tue, Mar 14, 2006 at 11:13:38PM -0500, Tom Lane wrote:

> Joe Sunday <sunday@csh.rit.edu> writes:
> > On Tue, Mar 14, 2006 at 03:56:51PM -0500, Tom Lane wrote:
> >> This error should result in dumping a list of per-context memory usage
> >> into the postmaster log ... could we have a look at that?
>
> > Here you go:
> > ...
> > ExecutorState: -346079744 total in 476747 blocks; 1498632 free (15 chunks); -347578376 used
>
> Ick, definitely something leaking in the ExecutorState context.  I can't
> guess what though.  Do you want to put together a self-contained test
> case for us to look at, or dig into it with gdb yourself?

It'll take me a while to come up with a dataset I can distribute that
causes it. My first pass at fake random data didn't seem to trigger it,
and I can't divulge the real data.

I can poke with gdb, is there a guide for what to look for somewhere?
I know my way around a debugger, but not the postgres internals all that
well.

--Joe

--
Joe Sunday <sunday@csh.rit.edu>  http://www.csh.rit.edu/~sunday/
Computer Science House, Rochester Inst. Of Technology