Hi,

I have installed and configured pound 2.0b6 on a Red Hat EL 4.0 server.  I 
was previously using the latest production release of pound, but wanted to 
try some of the new features available in the beta version, as i was 
having issues with a misbehaving Lotus Domino server.  The production 
release of pound worked perfectly, by the way, and I did not have any 
issues like the one I am about to describe

When I start pound (no daemon) and then access it via a web browser, the 
worker thread crashes and restarts every time I attempt to access.

[...] pound: MONITOR:  worker exited on signal 11, restarting

and occasionally, multiple failures, indicated by:

[...] last message repeated 4 times

pound says my config file is good, but here it is (I commented a bunch of 
stuff out in the process of troubleshooting the problem).

[root(at)camproxy01 ~]# cat /usr/local/etc/pound.cfg
#pound configuration file
#
#Global Directives
User "pound"
Group "pound"
# 0=foreground, 1=daemon
Daemon 0
# 0=none, 1=normal, 2=extended
LogLevel 2
Alive 30
ListenHTTP
        Address 192.152.136.25
        Port 80
        xHTTP 0
        WebDAV 0
        Client 10
#       Change30x 1
End

#ListenHTTPS
#       Address 192.152.136.25
#       Port 443
#       xHTTP 0
#       WebDAV 0
#       Client 10
#       Change30x 1
#       Cert "/usr/local/etc/poundcert/newcert.pem"
#       CAList "/usr/local/etc/poundcert/caroots.pem"
#       VerifyList "/usr/local/etc/poundcert/cacert.pem"
#End

#Service
#       URL ".*"
#       HeadRequire "Host:.*camproxy01.int.camsys.com.*"
#               BackEnd
#                       Address 192.152.136.11
#                       Port 80
#                       Priority 1
#               End
#               Session
#                       Type IP
#                       TTL 300
#               End
#
#End

Service
        BackEnd
                Address www.mit.edu
                Port 80
        End
End

Any assistance would be greatly appreciated!

--jeff

Jeffrey R. Brown
Cambridge Systematics, Inc.
100 CambridgePark Dr., Suite 400
Cambridge, MA 02140
tel  617 354 0167
fax  617 354 1542
jbrown(at)camsys.com
www.camsys.com

> Hi,[...]

I'm also seeing similar messages on CentOS 4.2/x86-64. 
I get the following

Jan 24 11:34:27 124 kernel: pound[20748]: segfault at 00000000000000d0
rip 0000003feaeacd9f rsp 00000000400bc700 error 6
Jan 24 11:34:27 124 pound: MONITOR: worker exited on signal 11,
restarting...

I've asked on amd64 specific list about this with no response though
this could be compiler/distro specific. I'll retry this with the
upcoming RHEL U3 release

On Tue, 2006-01-24 at 13:14 -0500, Jeffrey Brown wrote:[...]

This is very often a symptom of a bad mix of libraries (OpenSSL, libc,
etc). Others have reported it working on RH, so I doubt it is a major
Pound issue, unless you have a really unique combination of factors
nobody else came across. I suggest you have a look at the RPMs Simon
Matter graciously made available at

http://www.invoca.ch/pub/packages/pound/

and try again. Please let us know.[...]

Hi,

I will take a look at the RPM you mentioned, as well as the various 
libraries.

I asked one of our software programmers to take a look, and this is what 
he had to say (forgive my ignorance - I have no idea what this means):

"Just for fun I copied the Pound-2.0b6 and created a debug version so I 
could run it in gdb. 
It does look like its an NPTL issue as this line of code fails:
pound.c:549 if(pthread_create($thr, $attr, thr_http, (void *)arg)) {
I get a SIGSEGV in regexec(at)(at)GLIBC_2.3.4 () from /lib/tls/libc.so.6
I'm not a kernel developer but I can't see anything wrong with that.  "

Does this clarify anything?  :)

Again, thanks for your assistance.

--jeff




Robert Segall <roseg(at)apsis.ch> 
01/25/2006 04:55 AM
Please respond to
pound(at)apsis.ch


To
pound(at)apsis.ch
cc

Subject
Re: [Pound Mailing List] pound 2.0b6 crashes when accessed






On Tue, 2006-01-24 at 13:14 -0500, Jeffrey Brown wrote:[...]
I [...]
to [...]
[...]

This is very often a symptom of a bad mix of libraries (OpenSSL, libc,
etc). Others have reported it working on RH, so I doubt it is a major
Pound issue, unless you have a really unique combination of factors
nobody else came across. I suggest you have a look at the RPMs Simon
Matter graciously made available at

http://www.invoca.ch/pub/packages/pound/

and try again. Please let us know.[...]

On Wed, 2006-01-25 at 10:26 -0500, Jeffrey Brown wrote:[...]

Please do - and please let us know if it makes a difference.
[...]

Unfortunately no. All it says it that there was a problem with some
regexec call, within some thread. Typical bad information from a
debugger that is not threads-aware.

It would be helpful if you could at least tell us what request (and
response) the problem occurred with. Try sniffing the traffic between
the clients and Pound, and if possible also between Pound and the
back-ends. Additionally please try setting LogLevel to 3 and show us
your log just prior to the failure.[...]