httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jan Kratochvil <l...@jankratochvil.net>
Subject Re: httpd-2.x servertype inetd
Date Fri, 05 Aug 2005 08:39:02 GMT
Hi Nick,

On Fri, 05 Aug 2005 17:20:10 +0900, Nick Kew wrote:
> Jan Kratochvil wrote:
> > Hi,
> > 
> > is there still no httpd-1.3 "servertype inetd" style MPM for httpd-2.x?
> > 
> > It would be useful for low-cost "config-reloading" testing/staging daemon.
> 
> In what sense low-cost?

External people with mobile devices should be able to use 'production' or
'testing' server. The 'testing' server will not be used in common case, though.
Machine is unfortunately low on memory and currently this unused 'testing'
server eats 25844KB of memory (free(1),kill(1),free(1),subtract w/o buffers,
running minimal set - 1 listen httpd, 2 children httpd; mod_perl in use).


> Apache's high startup cost is self-reinforcing.  We know it's a once-only
> thing, so we have every module do expensive things at startup rather than
> per-request.  I don't see how inetd would affect that.

I care primarily of the not-in-use case. 'testing' server can get swapped out
but Linux kernel does not swap out inactive processes too much actively.

...
> > Do you see it as a separate MPM or some option to prefork.c would be enough?
> > There is already the "-X"/"DEBUG" hack.
> 
> Well, in principle I guess it could be an MPM.

It could be MPM but it could be also patch for prefork.c.
Asking what would get accepted to not to later have to code it second time.

 * patch con: Increased prefork.c conditions-complexity a bit.
 * MPM   con: Mostly duplicite code with prefork.c.
 * MPM   con: Increased codebase.


Thanks for reply,
Lace


Mime
View raw message