httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Slemko <ma...@worldgate.com>
Subject Re: process model again...
Date Wed, 08 Apr 1998 02:18:04 GMT
On Tue, 7 Apr 1998, Dean Gaudet wrote:

> So I've been continuing with trying to figure out if the current
> perform_idle_server_maintenance() code is right.  I'm now convinced that
> we no longer need MinSpareServer, but can't give a good mathematical
> reason for why.  It just doesn't seem necessary to keep many spare when
> the server will exponentially spawn if a load presents itself.

We still need MinSpareServer in theory because there is some delay x
involved in forking a new child when nrequests == nchildren, especially if
we only check the scoreboard once a second or whatever.  Your argument is
that this delay is so puny compared to everything else that there is no
point, right?

> [Tue Apr  7 09:40:47 1998] [info] server has 18 children, 0 are idle, spawning 1
> [Tue Apr  7 15:10:56 1998] [info] server has 19 children, 0 are idle, spawning 1
> [Tue Apr  7 15:10:58 1998] [info] server has 20 children, 0 are idle, spawning 1
> [Tue Apr  7 15:10:59 1998] [info] server has 21 children, 0 are idle, spawning 2
> [Tue Apr  7 15:11:22 1998] [info] server has 23 children, 21 are idle, killing one
> [Tue Apr  7 15:11:33 1998] [info] server has 22 children, 22 are idle, killing one

You go from 21 in use to 2 in use in a period of 30 seconds.  Even
if you remove minspareservers, you will still be yoyoing unless
you crank max high.  No matter how you do the pool, Max-Min has to
be > your average medium term connection change or you get killing
and starting.  I think this is just your server being very
freaky.  I don't see this to the same extent. 

Hmm.  Or do we want to tell children to stop listening without killing
them when we have lots spare?  Probably not worth it except in theory.


Mime
View raw message