Hi,

I also have a server with RedHat Advanced Server 3.0. After some 
searching I found an src rpm to install Pound (since comping doesn't 
work, also not with the 1.7 version, why not ??? I know why it doesn't 
compile, but why isn't it fixed in the 1.7 version ?). And it's working 
now, but it's very slow (just like the one who started this thread).
Here a small piece of my /var/log/messages:

Mar 25 17:24:32 s3 pound: 217.195.249.226 GET 
/campings/de_paardekreek?lng=nl HTTP/1.1 - HTTP/1.1 200 OK
Mar 25 17:24:32 s3 pound: MONITOR: worker exited nurmally 127, restarting...
Mar 25 17:24:35 s3 pound: 217.195.249.226 GET 
/campings/images/figuur-paardekreek.jpg HTTP/1.1 - HTTP/1.1 200 OK
Mar 25 17:24:35 s3 pound: 217.195.249.226 GET 
/campings/images/collage/paard-zwembad_paddestoel.jpg HTTP/1.1 - 
HTTP/1.1 200 OK
Mar 25 17:24:35 s3 pound: MONITOR: worker exited nurmally 127, restarting...
Mar 25 17:24:38 s3 pound: 217.195.249.226 GET 
/campings/images/collage/paard-watersport_banaan.jpg HTTP/1.1 - HTTP/1.1 
200 OK
Mar 25 17:24:38 s3 pound: MONITOR: worker exited nurmally 127, restarting...
Mar 25 17:24:42 s3 pound: 217.195.249.226 GET 
/campings/topics/campings/de_paardekreek/index/paard1f.jpg HTTP/1.1 - 
HTTP/1.1 200 OK
Mar 25 17:24:42 s3 pound: MONITOR: worker exited nurmally 127, restarting...

As you can see, it's onlu serving 1 file in 3 seconds (sometimes 2).
Any Ideas what is causing this ? I'm pretty sure it has nothing to do 
with the hardware, since it's just brand new, and the webservers are in 
a local network. And without pound, everything goes very fast.

My config:
------
User nobody
Group nobody
RootJail /usr/share/pound
ListenHTTP *,80

# Catch-all server(s)
UrlGroup ".*"
BackEnd 10.1.0.4,80,1
BackEnd 10.1.0.5,80,1
Session IP 300
EndGroup
-----

Any Idea for solving this problem ?

Regards,

Sandy[...]

If RHEL3 is NPTL-based like RH9 and Fedora I suggest trying the
LD_ASSUME_KERNEL trick:

$ LD_ASSUME_KERNEL=2.4.19 /usr/bin/pound

or try commenting out RootJail.

Sean

On Thu, 2004-03-25 at 17:34, Sandy Pleyte wrote:[...]

Telling us what the compilation problems are would be helpful - I was
not aware that RH AS3 has _any_ compilation issues.
[...]

Your worker process dies with just about every request or two. This
makes it naturally slow - it needs to restart on every request.

Very often this is a symptom of a wrong OpenSSL library - for example
one without threads support. I also suspect that the RH implementation
of NPTL still has major issues - see other messages in the archives on
this subject.[...]

Robert Segall wrote:
[...]

The problem starts with the './configure --with-ssl=/usr/include', it 
seems the it can't find the krb5.h (which is located in
/usr/kerberos/include/krb5.h)

configure:3149: checking openssl/ssl.h usability
configure:3162: gcc -c -g -O2 -DAEMON -I/usr/include/include 
-I%{_prefix}/kerberos/include -D_REENTRANT -Wstrict-prototypes -pipe 
conftest.c >&5
In file included from /usr/include/openssl/ssl.h:179,
                  from configure:3216:
/usr/include/openssl/kssl.h:72:18: krb5.h: No such file or directory
In file included from /usr/include/openssl/ssl.h:179,
                  from configure:3216:

On some earlier posts, Simon Matter <simon.matter(at)invoca.ch> posted a 
url (http://www.invoca.ch/pub/packages/pound/)
where you can find 
src.rpm. This one does compile on RH AS3. And according to an other post 
the magic should be in this piece of code (from the pound.spec file):

%build
# kerberos include is needed (because of openssl-0.9.7 ?)
CPPFLAGS="-I%{_prefix}/kerberos/include"
export CPPFLAGS
%{configure} \
%if %{enablemsdav}
   --enable-msdav \
%endif
%if %{enableunsafe}
   --enable-unsafe \
%endif
   --sysconfdir=%{_sysconfdir}/%{name}
%{__make}

I don't know much of makefiles and .spec files, but I'm correct the 
CPPFLAGS are the solution ? If so, what does it need to change the 
makefile so it also compiles ?
[...]
Well, commenting out the RootJail was the solution, it now goes a lot 
faster. Maybe this is something for the FAQ ?

Regards,

Sandy[...]

On Fri, 2004-03-26 at 09:15, Sandy Pleyte wrote:[...]

The --with-ssl=/usr/include is not correct. You should not need any flag
here.
[...]

Try it without any flag to get rid of the %{_prefix} nonsense.
[...]

This has been discussed to death here. Please try compiling Pound
normally and run it with the old thread model.[...]

Robert Segall wrote:[...][...]
[...][...]
Ok, this the config.log of ./configure without any flag (version 1.7).

configure:3149: checking openssl/ssl.h usability
configure:3162: gcc -c -g -O2 -DAEMON  -D_REENTRANT -Wstrict-prototypes 
-pipe conftest.c >&5
In file included from /usr/include/openssl/ssl.h:179,
                  from configure:3216:
/usr/include/openssl/kssl.h:72:18: krb5.h: No such file or directory
In file included from /usr/include/openssl/ssl.h:179,
                  from configure:3216:

It still can't find the krb5.h.

Regards,

Sandy
[...]

On Fri, 2004-03-26 at 15:22, Sandy Pleyte wrote:[...]

Check which RPM includes it and install it (I vaguely recall something
about Kerberos development libraries). You may want to let RedHat know
about this bad dependency (why is openssl-devel installed without
krb-devel).[...]

Robert Segall wrote:
[...][...][...]

I allready have both packages:

openssl-devel-0.9.7a-33.4
krb5-devel-1.2.7-19

Also the file does exist on my system:

/usr/kerberos/include/krb5.h

Only the ./configure script can't find it.

Regards,

Sandy[...]

On Fri, 2004-03-26 at 16:53, Sandy Pleyte wrote:[...]

You may want to update that - 0.9.7a had several vulnerabilities which
have been corrected since.
[...]

This is nothing to do with configure: all it does is to test if ssl.h is
usable. krb.h is apparently included by ssl.h - configure just reports
an error. Try compiling some SSL-dependent program of your own (such as
the examples in the OpenSSL) to see that.

Again, I suggest you check with RedHat if a fix exists for the problem.
If not try to add (manually) -I/usr/kerberos/include to your CFLAGS. BTW
- shouldn't that be symlinked to /usr/include/kerberos?

Any RedHat users here who would care to help?[...]

I get this problem all the time when compiling ssl-aware software
on Redhat.

Robert is correct - the easiest solution is to add -I/usr/kerberos/include
to your CFLAGS -- ssl.h includes kssl.h which includes krb5.h.
(I haven't built Pound 1.7 yet, but I'm guessing this should work)

-jason

On Fri, Mar 26, 2004 at 05:03:10PM +0100, Robert Segall wrote:[...]
[...]

Hi,

thanks for you answer, but I'm not a programmer, so any idea what file I 
need to edit ?
I just did 'export CFLAGS=-I/usr/kerberos/include' and then ./configure, 
but it gives me a warning now:

configure: WARNING: openssl/ssl.h: accepted by the compiler, rejected by 
the preprocessor!
configure: WARNING: openssl/ssl.h: proceeding with the preprocessor's result
configure: WARNING:     ## ------------------------------------ ##
configure: WARNING:     ## Report this to bug-autoconf(at)gnu.org. ##
configure: WARNING:     ## ------------------------------------ ##

And the config.log says:

configure:3149: checking openssl/ssl.h usability
configure:3162: gcc -c -I/usr/kerberos/include -DAEMON  -D_REENTRANT 
-Wstrict-prototypes -pipe conftest.c >&5
configure:3165: $? = 0
configure:3168: test -s conftest.o
configure:3171: $? = 0
configure:3181: result: yes
configure:3185: checking openssl/ssl.h presence
configure:3196: gcc -E  -D_REENTRANT -Wstrict-prototypes -pipe conftest.c
In file included from /usr/include/openssl/ssl.h:179,
                  from configure:3218:
/usr/include/openssl/kssl.h:72:18: krb5.h: No such file or directory
configure:3202: $? = 1

It seems it only uses the CFLAG for the first gcc, but not for the 
second one ???

Regards,

Sandy


Jason Vasquez wrote:
[...][...]
>>>
>>>openssl-devel-0.9.7a-33.4[...]
>>>krb5-devel-1.2.7-19
>>>
>>>Also the file does exist on my system:
>>>
>>>/usr/kerberos/include/krb5.h
>>>
>>>Only the ./configure script can't find it.[...][...]

[...]

On Mon, 2004-03-29 at 09:07, Sandy Pleyte wrote:[...]

Try CPPFLAGS - that might help.[...]

Robert Segall wrote:[...]

Thanks, it works, I can compile it, and it runs nice till now :)

Regards,

Sandy[...]