httpd-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Arthur Kreitman" <a...@congruent.com>
Subject RE: [users@httpd] Tuning the MaxKeepAliveRequests and KeepAliveTimeout http.conf parameters
Date Mon, 20 Nov 2006 20:57:00 GMT
And what about bad side effects from increasing the linger time?

> -----Original Message-----
> From: William A. Rowe, Jr. [mailto:wrowe@rowe-clan.net]
> Sent: Monday, November 20, 2006 12:50 PM
> To: users@httpd.apache.org
> Subject: Re: [users@httpd] Tuning the MaxKeepAliveRequests and
> KeepAliveTimeout http.conf parameters
> 
> Arthur Kreitman wrote:
> >
> > I noticed a random but all to frequent (perhaps 1 or 2 or 3 for
every
> > few hundred http requests) error from ap_core_output_filter (about
line
> > 903 in core_filters.c).   The error is an os 10054 "An existing
> > connection was forcibly closed by the remote host".   At the same
time,
> > the client side would believe the server performed a premature
socket
> > close.
> >
> > Regardless of the value of KeepAliveTimeout, if I set
> > MaxKeepAliveRequests to 1 (ie no keep alive) the error occurs for
almost
> > 80% or 90% of the http requests
> >
> > If I don't set MaxKeepAliveRequests, which then uses the default
value,
> > by setting KeepAliveTimeout between 60 and 90 seconds, the error
> disappears.
> 
> DING DING DING - thank you for providing the essentials for a
reproducible
> test case!!!
> 
> It sounds like we are not handling keep alive disconnection in the
best
> manner possible, leading to confusion by the ISAPI application.
> 
> > My understanding of KeepAliveTimeout is that it plays with the
socket
> > linger option.   What are the side effects of setting
KeepAliveTimeout
> > to a high value?   What are the side effects of setting
> > MaxKeepAliveRequests to a high value?
> 
> 1. fewer workers available to handle high load
> 
> 2. more exposure to denial of service (every daemon has an implicit
denial
>    of service that it can handle only X parallel requests over N
amount of
>    time - keeping N low helps mitigate this).
> 
> Contra wise, the benefit is that the next request from the same client
is
> handled more quickly, if you assume that 'next page' requests will
happen
> within the keepalivetimeout period.
> 
> 
> ---------------------------------------------------------------------
> The official User-To-User support forum of the Apache HTTP Server
Project.
> See <URL:http://httpd.apache.org/userslist.html> for more info.
> To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
>    "   from the digest: users-digest-unsubscribe@httpd.apache.org
> For additional commands, e-mail: users-help@httpd.apache.org


---------------------------------------------------------------------
The official User-To-User support forum of the Apache HTTP Server Project.
See <URL:http://httpd.apache.org/userslist.html> for more info.
To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
   "   from the digest: users-digest-unsubscribe@httpd.apache.org
For additional commands, e-mail: users-help@httpd.apache.org


Mime
View raw message