On Wed, Jan 08, 2025 at 02:48:10PM +0100, Frédéric Yhuel wrote: > For what it's worth, although I would have preferred the sub-linear growth > thing, I'd much rather have this than nothing.
+1, this is how I feel, too. But I also don't want to add something that folks won't find useful.
> And I have to admit that the proposed formulas were either too convoluted or > wrong. > > This very patch is more straightforward. Please let me know if I can help > and how.
I read through the thread from the top, and it does seem like there is reasonably strong support for the hard cap. Upon a closer review of the patch, I noticed that the relopt was defined such that you couldn't disable autovacuum_max_threshold on a per-table basis, so I fixed that in v4.
-- nathan
nathan,
Please also provide the tests on the new parameter you want to introduce.
