httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <>
Subject Re: [PATCH] lingering_close performance improvement
Date Mon, 10 Feb 1997 18:19:38 GMT
Under 1.1.1 I saw moderate FIN_WAIT_2s w/keepalive enabled (although I
always had the mozilla/2 hack in) -- about 20 times the numbers of hits/s. 
When I disabled keepalive it would drop to "background noise" (about 10
times the number of hits/s).

Under 1.2b7-dev I've only let the FIN_WAIT_2s get up into the 25-30 times
the hits/s range before rebuilding with -DNO_LINGCLOSE when it dropped
back down to 20 times.

I can retry with one of the servers running with lingering_close() and see
what happens.  But I won't get to this until late this week. 


On Mon, 10 Feb 1997, Jim Jagielski wrote:

> Roy T. Fielding wrote:
> > 
> > The ONLY reason we are seeing this problem now, and not
> > before, is because keep-alive clients behave differently than non-keep-alive
> > clients.  The FIN_WAIT_2 timeout is necessary because it will cause an
> > Apache 1.1.3 server (with full keep-alive enabled, including Mozilla/2)
> > to reboot just as quickly as an Apache 1.2b6 server.  We fixed the other
> > problems with lingering_close after 1.2b4, and we can reduce the overhead
> > even more with my performance patch, but we can't do a damn thing about
> > the client keeping a connection alive when the server-side has closed.
> > 
> I seem to recall that there were many(?) cases where if people recompiled
> with NO_LINGCLOSE the FIN_WAIT_2 problem "went away"... Now this may
> have been fixed with the latest rev to l_c(), but maybe not. In any
> case, no other changes were done; just using l_c() or not. I seem
> to recall that HotWired servers were hit in this way... Does the
> latest version of l_c() remove the problem? If so, then that's cool
> (although there still seems to be some disagreement on whether
> l_c() is even needed according to the 1.1 specs) but if not then
> we can't blame it on keepalive clients.
> -- 
> ====================================================================
>       Jim Jagielski            |       jaguNET Access Services
>           |
>                   "Not the Craw... the CRAW!"

View raw message