CVS of todays code freezing erratically ..

[Anthony Brock <abrock@georgefox.edu>]

Okay,

I am using the development branch for some basic services here on 
campus.  For the most part, it has worked perfectly with no 
glitches.  Excellent work!

However, today I started encountering some strange behavior.  Basically, I 
was running a version against the CVS code of a couple weeks ago (post beta 
release 2.0), and this morning the server stopped responding.  I telneted 
into the Sun Sparc 250 with 2gb of ram running Solaris 2.8, and everything 
appeared to be normal.  The process was running, cpu load was very small 
(0.03), and everything else was responding.  However, the LDAP server was 
not responding.  I tried kill the process to restart, and it would not exit 
(slapd).  The slurpd daemon exited smoothly.  I finally killed it with a:

kill -9 <pid>

I restarted the machine, and everything appeared to be happy again.  I went 
on to other projects since this seemed a one-time glitch after running for 
a couple weeks.  Then at noon, the same thing happened.  This time, I did:

netstat -na

I was astonished at the number of connections that were associated with the 
389 port (see attached file) (okay, don't see attached file, since it 
bounced, available at http://www.georgefox.edu/netstat.txt).  I am since 
been able to reproduce this on a consistent basis.  I have a CGI perl 
script for internal use only which results in several hundred queries being 
run against the LDAP server.  It can, in a minute or two (all I have to do 
is run it 20 or 30 times), freeze the server.  Once frozen, it will only 
exit via a 'kill -9 <pid>'.

I have updated against the current CVS (2:30 PM PDT), and the same problem 
exists.

Could you give me some pointers as to what would cause this or how to 
better troubleshoot?  This is the first problem I have been able to 
reproduce while running '/opt/LDAP/libexec/slapd -f 
/opt/LDAP/etc/openldap/slapd.conf -d 8'.  I can easily produce any traces 
you wish to see.  The perl script is running under an Apache web server, 
and the perl LDAP modules were compiled against LDAP 1.2.9.  The Apache web 
server is using a custom build module adapted from auth_ldap.c for our 
uses, and works perfectly against our 1.2.11 server.  It was built against 
the same libraries as the perl modules.

The LDAP beta server was compiled with:

./configure --prefix=/opt/ldap --with-ldbm-api=berkely --with-threads

Thanks in advance,

Tony
******************************************************************************
* Anthony Brock                                         abrock@georgefox.edu *
* Director of Network Services                         George Fox University *
******************************************************************************