headers
Xose Vazquez Perez | 1 Jun 2003 16:41
Picon

Re: Tux Patch for Linux-2.4.21-rc1

Matthew Hodgson wrote:

>On Wed, 30 Apr 2003, Sascha Schumann wrote:
>>     Hi,
>>
>>     here is an updated patch for kernel 2.4.21-rc1.
>>
>>     http://schumann.cx/tux2-full-2.4.21-rc1-final-A3.bz2
>>
>>     The load and response times have improved significantly
>>     on www.php.net thanks to Tux.
>>
>>     - Sascha

>It's great to see that someone's updating the TUX kernel patches for
>current kernels.  By any chance have you encountered/solved the "system
>hang when logging enabled" bug that has rendered TUX unusable for many of
>its users, during your PHP/patch-updating travels?

latest tux release, plus bugzilla fixes, are inside redhat kernels too.
kernel: 2.4.20-13.7 -> tux: linux-2.4.15-tux2.patch
and user-space tux-2.2.10 are at rawhide

regards,
--

-- 
Software is like sex, it's better when it's bug free.
Permalink | Reply | 1 comment |
headers
David A Sinck | 6 Jun 2003 01:21

ip help


I'm attempting to 'properly' set up tux to listen to a single IP
hosted on the box, and while I can just 

echo 1.2.3.4 > /proc/net/tux/0/listen/0

in rc.local, it seems so unclean when there's /etc/sysctl.tux that's
running around and is useful for net.tux.virtual_server = 1 and stuff.

Grepping the mailing list archives didn't produce any useful hits nor
did rtfm nor stfw.

Any hints?  

TIA,

David
Permalink | Reply | 0 comments |
headers
Xuan Chen | 8 Jun 2003 18:57
Picon
Favicon

error message when sending a web request to tux

I are doing some experiment with my web server. Basically, I wrote a 
program to send out url to the server and receive responses---just like 
other web traffic generator (webstone, surge, etc). It works fine with 
Apache, ie, I can see the web pages retrieved. 

But, when I switched to tux yesterday. I kept getting error messages 
complaining: 

Error on Read: Connection reset by peer

It seems when I do recv, it always returns -1. 

I also checked tux log on the server, it show the request has been served 
correctly, ie, status 200. 

Could somebody give me a hint on how I should modify my program so that it 
works with tux also? Thanks.

btw, I suspect that I should have written my program with asynch socket, 
cz it looks to me that tux is using asynch I/O. Is that the problem? 
Thanks.

cheers,
-chen

--

-- 
Xuan Chen
USC/ISI
Permalink | Reply | 1 comment |
headers
Xuan Chen | 9 Jun 2003 19:21
Picon
Favicon

Re: error message when sending a web request to tux

I just found that if I use HTTP 1.1 instead of HTTP 1.0 in the web 
requests, my program works fine. So, my follow-up question is "is tux 
compatible with HTTP 1.0?" Or, maybe, I still miss something in my 
program... 

On Sun, 8 Jun 2003, Xuan Chen wrote:

> I are doing some experiment with my web server. Basically, I wrote a 
> program to send out url to the server and receive responses---just like 
> other web traffic generator (webstone, surge, etc). It works fine with 
> Apache, ie, I can see the web pages retrieved. 
> 
> But, when I switched to tux yesterday. I kept getting error messages 
> complaining: 
> 
> Error on Read: Connection reset by peer
> 
> It seems when I do recv, it always returns -1. 
> 
> I also checked tux log on the server, it show the request has been served 
> correctly, ie, status 200. 
> 
> Could somebody give me a hint on how I should modify my program so that it 
> works with tux also? Thanks.
> 
> btw, I suspect that I should have written my program with asynch socket, 
> cz it looks to me that tux is using asynch I/O. Is that the problem? 
> Thanks.
> 
> cheers,
(Continue reading)

Permalink | Reply | 0 comments |
headers
Shane Allen | 10 Jun 2003 22:01

Re: Tux Patch for Linux-2.4.21-rc1

On Sun, Jun 01, 2003 at 04:41:35PM +0200, Xose Vazquez Perez wrote:
> Matthew Hodgson wrote:
> 
> >current kernels.  By any chance have you encountered/solved the "system
> >hang when logging enabled" bug that has rendered TUX unusable for many of
> >its users, during your PHP/patch-updating travels?
> 
> latest tux release, plus bugzilla fixes, are inside redhat kernels too.
> kernel: 2.4.20-13.7 -> tux: linux-2.4.15-tux2.patch
> and user-space tux-2.2.10 are at rawhide

This post made me wonder what's going on with tux. We haven't run into
any issues (of course, we're not logging), but I didn't remember any
sort of discussion centered around new releases within, say the past
year or so.

I hit http://people.redhat.com/mingo/TUX-patches/ and see that the
userspace modules in there are version 2.2.7 (5/5/02), and the latest
kernel patch is 2.4.18-A3 (2/4/02)...

Why then is rawhide using 2.4.15, and where did 2.2.10 user-space utils
come from?

For reference, we're using currently 2.4.17-A1 kernel patch and 2.2.3
user-space.

Thoughts?

--

-- 
Shane Allen <shane <at> sell.com>
(Continue reading)

Permalink | Reply | 0 comments |
headers
Simon Brock | 12 Jun 2003 16:41
Picon

Re: Tux Patch for Linux-2.4.21-rc1

On a related theme, where should we send bug fixes?

We have found (and fixed) a couple of bugs in the HTTP protocol 
implementation.  I can send the fixes here but they should probably 
go elsewhere as well.  What do you think I should do with them?

The bugs are related to how URLs like this are handled:

http://www.site.com/directory

when directory is a directory name.  A web server should redirect to:

http://www.site.com/directory/

but instead TUX redirects to:

http://site.com/directory/

(or worse if you are using stripping).  Also, if you are running on a 
non-standard port, it forgets it from the redirect e.g.

http://www.site.com:1234/directory

becomes:

http://site.com/directory/

Regards,
Simon.
(Continue reading)

Permalink | Reply | 7 comments |
headers
Chris Davies | 12 Jun 2003 16:49
Favicon

Re: Tux Patch for Linux-2.4.21-rc1

another bug that we fixed was during the host checking.

Tux would close the connection if you were in virtual mode 3 and
requested a domain that had not been set up.

http://domain.com/

but

/var/www/1.2.3.4/domain.com 

didn't exist.  We changed it to pass to the backend rather than kill the
connection.

I'd say, at least post the diff to the list -- I don't know that anyone
is really maintaining things.

I'll get the diff for our change and post it as well.

On Thu, 2003-06-12 at 10:41, Simon Brock wrote:
> On a related theme, where should we send bug fixes?
> 
> We have found (and fixed) a couple of bugs in the HTTP protocol 
> implementation.  I can send the fixes here but they should probably 
> go elsewhere as well.  What do you think I should do with them?
Permalink | Reply | 3 comments |
headers
Matthew Hodgson | 14 Jun 2003 04:34

Re: Tux Patch for Linux-2.4.21-rc1

On Thu, 12 Jun 2003, Simon Brock wrote:

> Regards,
> Simon.
>
> PS Just for the record, we are running TUX from the RH8.0 and are
> logging traffic via tux.  What is the problem with logging?

Several people (myself included) had an issue where TUX would completely
hang a server after a few (1-10) days, yielding it unresponsive other than
to ping.  Finding any common factor between people experiencing the
problem was dubious - in the end the only thing that could be obviously
seen in common was the fact that all the sufferers appeared to have
logging enabled.  For my purposes, despite serving a fairly steady 50-70
hits/s off TUX, logging is essential for checking referrers and popularity
of parts of sites - and thus turning it off wasn't really an option.

The whole subject was covered pretty comprehensively in the thread "random
system crashes", which can be browsed at MARC at
http://marc.theaimsgroup.com/?l=tux-list&m=103460993003287&w=2 as various
people on the list compared & contrasted their setups (operational and
otherwise) to try to diagnose the problem.
http://marc.theaimsgroup.com/?l=tux-list&m=103472615516863&w=2 was my
chain of thought from wrestling with TUX back in May of last year, fwiw.

http://marc.theaimsgroup.com/?l=tux-list&m=102260919428123&w=2 is a report
of the problem also from May, back when people thought it might be an
issue with manipulating log files whilst TUX was writing to them.

If anyone ever found a solution i'd be very interested - thttpd is getting
(Continue reading)

Permalink | Reply | 2 comments |
headers
Max Helmet | 16 Jun 2003 12:12
Picon

Re: Tux Patch for Linux-2.4.21-rc1

> On Thu, 2003-06-12 at 10:41, Simon Brock wrote:
> > On a related theme, where should we send bug fixes?

On Thu, 2003-06-12 at 16:49, Chris Davies wrote:
> I'd say, at least post the diff to the list -- I don't know that anyone
> is really maintaining things.
> I'll get the diff for our change and post it as well.

Thank you for doing this!

Unfortunately Ingo is occupied with many other interesting problems and it
seems as if he currently doesn't have the time to develop Tux further.
Nonetheless it's still one of the most advanced webservers with some key
advantages over thttpd, fnord and others. Beyond its unbeaten speed, it has
the ability to run "in front of" Apache on the same IP and port, eg. as an
web-accellerator, and the possibility to extend its functionality with
user-loadable modules. As I see it, this combination can make it a very
powerful solution for chat-servers as well...

At present, this mailing-list is the only way to share patches, installation
hints and other resources. However, may be someday a small group of
webmasters will gather together for the purpose of coordinating their
changes. It wouldn't be the first time in history... (see
http://httpd.apache.org/ABOUT_APACHE.html#How)

Regards,
Maximilian
Permalink | Reply | 2 comments |
headers
Michael T. Halligan | 16 Jun 2003 12:31
Gravatar

Re: Tux Patch for Linux-2.4.21-rc1

You might want to consider an alternative.

Check out fnord..

http://www.fefe.de/fnord/

It might not have all the functionality, but it's a blazingly fast static
content server, and it has an active support community behind it. Tux is
pretty stagnant right now, and I've personally removed it from production
for all of my clients due to that.   I'm also told that BOA is pretty
competitive in speed. 

On Mon, 16 Jun 2003, Max Helmet wrote:

= > On Thu, 2003-06-12 at 10:41, Simon Brock wrote:
= > > On a related theme, where should we send bug fixes?
= 
= On Thu, 2003-06-12 at 16:49, Chris Davies wrote:
= > I'd say, at least post the diff to the list -- I don't know that anyone
= > is really maintaining things.
= > I'll get the diff for our change and post it as well.
= 
= Thank you for doing this!
= 
= Unfortunately Ingo is occupied with many other interesting problems and it
= seems as if he currently doesn't have the time to develop Tux further.
= Nonetheless it's still one of the most advanced webservers with some key
= advantages over thttpd, fnord and others. Beyond its unbeaten speed, it has
= the ability to run "in front of" Apache on the same IP and port, eg. as an
= web-accellerator, and the possibility to extend its functionality with
(Continue reading)

Permalink | Reply | 1 comment |

Gmane