Re: Time to increase hash_mem_multiplier default?

2022-02-16 Thread Peter Geoghegan
On Tue, Feb 15, 2022 at 8:17 AM Justin Pryzby wrote: > The only reason not to is that a single-node hash-aggregate plan will now use > 2x work_mem. Which won't make sense to someone who doesn't deal with > complicated plans (and who doesn't know that work_mem is per-node and can be > used

Re: Time to increase hash_mem_multiplier default?

2022-02-15 Thread Justin Pryzby
On Mon, Feb 14, 2022 at 10:32:43PM -0800, Peter Geoghegan wrote: > On Wed, Jan 19, 2022 at 11:32 AM John Naylor > wrote: > > I don't have anything really profound to say here, but in the last > > year I did on a couple occasions recommend clients to raise > > hash_mem_multiplier to 2.0 to fix

Re: Time to increase hash_mem_multiplier default?

2022-02-14 Thread Peter Geoghegan
On Wed, Jan 19, 2022 at 11:32 AM John Naylor wrote: > I don't have anything really profound to say here, but in the last > year I did on a couple occasions recommend clients to raise > hash_mem_multiplier to 2.0 to fix performance problems. I would like to push ahead with an increase in the

Re: Time to increase hash_mem_multiplier default?

2022-01-19 Thread John Naylor
On Sun, Jan 16, 2022 at 7:28 PM Peter Geoghegan wrote: > > The current hash_mem_multiplier default is 1.0, which is a fairly > conservative default: it preserves the historic behavior, which is > that hash-based executor nodes receive the same work_mem budget as > sort-based nodes. I propose that

Time to increase hash_mem_multiplier default?

2022-01-16 Thread Peter Geoghegan
The current hash_mem_multiplier default is 1.0, which is a fairly conservative default: it preserves the historic behavior, which is that hash-based executor nodes receive the same work_mem budget as sort-based nodes. I propose that the default be increased to 2.0 for Postgres 15. Arguments in