httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeff Trawick <traw...@attglobal.net>
Subject Re: [PATCH] threaded.c idle_spawn_rate gets huge relative to ap_daemons_limit
Date Wed, 25 Jul 2001 20:15:10 GMT
Greg Ames <gregames@remulak.net> writes:

> Jeff Trawick wrote:
> > 
> > Even if child process creation is limited by ap_daemons_limit, we
> > still write trace messages saying we're about to spawn x children,
> > where x > ap_daemons_limit.
> > 
> > This patch doesn't let idle_spawn_rate get over ap_daemons_limit / 2.
> > It writes the log message at that threshold as well.
> > 
> > @@ -1030,7 +1032,7 @@
> >         }
> >         else {
> > 
> > -           if (idle_spawn_rate >= 8) {
> > +           if (idle_spawn_rate >= ap_daemons_limit / 2) {
> >                 ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_INFO, 0, ap_server_conf,
> >                              "server seems busy, (you may need "
> >                              "to increase StartServers, ThreadsPerChild "
> 
> ...but then we can't see it ramp up.  Do we need to?  dunno. I'd prefer
> to see it ramp up until we're happy with how it works.
> 
> I realize we don't want to put out a log msg with wrong info.  Looks
> like "free_length" is what we should be using in this msg, rather than
> idle_spawn_rate. 

yessir, free_length looks like the field to trace, and we should do so
when this condition occurs

    if (free_length >= 8) {
    }

rather than when this one occurs

    if (idle_spawn_rate >= 8) {
    }

(good news: can't recreate any threaded.c weirdness on AIX or Solaris,
though certainly I'm not running close to user/system resource limits
on those boxes, unlike with the Linux box where I have troubles when I
pound threaded.c)

-- 
Jeff Trawick | trawick@attglobal.net | PGP public key at web site:
       http://www.geocities.com/SiliconValley/Park/9289/
             Born in Roswell... married an alien...

Mime
View raw message