On Thu, Jul 22, 2021 at 9:53 AM Peter Geoghegan <pg@bowt.ie> wrote:
> I suspect David's theory about hash_agg_set_limits()'s ngroup limit is
> correct. It certainly seems like a good starting point.
I also suspect that if Laurent set work_mem and/or hash_mem_multiplier
*extremely* aggressively, then eventually the hash agg would be
in-memory. And without actually using all that much memory.
I'm not suggesting that that is a sensible resolution to Laurent's
complaint. I'm just pointing out that it's probably not fundamentally
impossible to make the hash agg avoid spilling through tuning these
GUCs. At least I see no evidence of that right now.
--
Peter Geoghegan