On 11/08/2012 11:35 PM, Carlos Henrique Reimer wrote:
> Hi Craig,
>
> work_mem is defined with 10MB and yes, there are triggers defined on
> both tables
Come to think of it, the triggers don't make any difference to memory
use for a SELECT anyway.
Your work_mem is perfectly reasonable.
The plan looks pretty harsh, with that big nested loop, but I'm not
aware of anything that'd cause that to run out of memory.
Personally I'd be attaching a debugger to it and seeing what it was
doing while it ran. I'm sure there are smarter ways to trace where the
memory use is going, though.
--
Craig Ringer
