httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Slemko <ma...@znep.com>
Subject Re: Keepalive and IE (was: Re: 5 forwarded messages on FIN_WAIT_2)
Date Sat, 18 Jan 1997 03:53:31 GMT
On Fri, 17 Jan 1997, Chuck Murcko wrote:

> Marc Slemko liltingly intones:
> > 
> > On Fri, 17 Jan 1997, Rob Hartill wrote:
> > 
> > > On 17 Jan 1997, James H Cloos Jr wrote:
> > > 
> > > > This reminds me of something.
> > > > 
> > > > When I first upgraded our server to 1.2, I forgot to add the line to
> > > > disable keepalives for Mozilla/2, but I did increase the keepalive
> > > > timeout from ~5s to 30s.  I later had to reduce that to 20s to keep
> > > > the number of children needed under 256.  The net effect was that we
> > > > finally had sessions where multiple files were sent.  As many as 20 or
> > > > 30 would occasionally showup, as reported by server-status.
> > > 
> > > Has anyone done any tests on the effects of varying the keepalive timeout
> > > value ?. It could make interesting reading.
> > > 
> > > I've always had mine set 4 seconds.
> > 
> > It will screw with the way lingering_close does things by changing the
> > timeout that it uses.  Set it to an odd value, and lingering_close will
> > behave quite differently. (Yes, this is another hint that I would still
> > like to know why lingering_close uses the keepalive timeout... if no one
> > responds I am tempted to think it is because no one knows because it
> > shouldn't...) 
> > 
> No one here knows because the guy who wrote it doesn't hang around the
> list anymore, and no one else has been interested in it until now.
> Perhaps it saved having yet another timeout variable. Frankly, I think
> it's a different problem than FIN_WAIT_2s.

I think that in the _general_ case it is entirely unrelated, however
remember we are seeing at least two different problems related to
FIN_WAIT_2 and possibly more.  If someone set the keepalive timeout very
low or high... something to keep in mind when looking at the problem and
perhaps clean up after 1.2. 



Mime
View raw message