Re: Massive memory use for star query

On 16/04/11 04:43, Tom Lane wrote:
> Mark Kirkwood<mark(dot)kirkwood(at)catalyst(dot)net(dot)nz> writes:
>> I've recently seen examples of star-like queries using vast amounts of
>> memory in one of our production systems. Here's a simplified example
>> using synthetic data (see attached to generate if desired):
>> SET geqo_threshold = 14;
>> SET from_collapse_limit = 14;
>> SET join_collapse_limit = 14;
> Well, if you're going to do the above, you should be expecting the
> planner to eat a lot of memory. There is a reason why the default
> values of those parameters are significantly lower than that ...
>

Ok - so with the settings at their defaults geqo chooses a semi-random
plan, and at least one of those (for the production variant of this
query anyway) eat massive (>100G) amounts of temp space - not really a
suitable outcome either.

I guess you have answered my first question - i.e yes this should eat
massive amount of ram as written - however are you sure there is no
memory leaking going on here?

regards

Mark