Tom Lane wrote:
> "Magnus Hagander" <magnus@hagander.net> writes:
>>> How about we just emit a warning..
>>>
>>> WARNING: Connections above 250 on Windows platforms may have
>>> unpredictable results.
>
>> That's probably a better idea. I'll go look at that unless people feel we should just stick it in docd/faq?
>
> Unless we've got some credible basis for citing a particular number,
> I don't think this will help much.
ok. Maybe a note in the docs or FAQ at least?
> Rainer Bauer <usenet@munnin.com> writes:
>> My guess is that Windows is running out of handles. Each backend uses about
>> 150 handles. 100 Backends means 15000 handles. Depending how many other
>> programs are currently running the no. of startable backends will vary
>> depending on the total handle limit Windows imposes.
>
> I find this theory very interesting; for one thing it explains the
> reported variability of results, since the non-Postgres demand for
> handles could be anything. Is there any way we could check it?
> If it's accurate, what we ought to be whining about is some
> combination of max_connections and max_files_per_process, rather
> than only considering the former.
It's not that simple. Merlin ran some checks, and drastically reducing
max_files_per_process made no measurable difference.
My best guess is it's due to the non-paged pool. Handles are a part of
what goes in there, but only a part.
//Magnus
