Re: [exim] 4096 connection barrier

Graeme Fowler wrote:
> On Mon, 2009-02-23 at 11:45 +0000, Ian Eiloart wrote:
>
>> So, Exim's limit isn't reached on my machines, but nevertheless the limit
>> needs revisiting for those who don't have my problem. Marc's right. The
>> limit is too low for modern hardware.
>>
>
> ...to which the followup question is: what is a suitable limit for
> modern hardware and OS combinations?
>
> To expand a little on your OSX limits, should Exim be checking on OSX
> that it never spawns more than 2499 processes (queue runners, delivery
> processes, inbound handlers and so on)? Should it factor OS variations
> at all, or should they simply be documented?
>
> What should be done to compare, say, Postfix / Sendmail et al?
>
> Most of these are (IMO) fairly rhetorical questions based on the fact
> that the following one-line change moves the goalposts:
>
> --- daemon.c.orig 2009-02-23 12:08:25.000000000 +0000
> +++ daemon.c 2009-02-23 12:08:40.000000000 +0000
> @@ -1199,3 +1199,3 @@
>
> - if (smtp_accept_max > 4095) smtp_accept_max = 4096;
> + if (smtp_accept_max > 16383) smtp_accept_max = 16384;
>
> But the first question stands: what value should be chosen? Or should
> there simply *not be* an upper limit, so people can skewer themselves if
> they choose an insane value for smtp_accept_max?
>
> Graeme
>
>
Ideally there should be no limit. Or - there should be a user setting to
cap it with 0 meaning no limit. But 16k would be a significant improvement.