Apache-SSL Weirdness on Tru64 5.0a

From: Scott Adkins <adkinss+lists.tru64-unix-managers_at_oak.cats.ohiou.edu>
Date: Fri, 04 Aug 2000 00:02:20 -0400

Well, I feel that I have exhausted my research and contacts with regards
to getting Apache-SSL working on our 5.0a system. Here is what I have:

  Tru64 Cluster 5.0a
  Apache 1.3.12
  Apache-SSL 1.3.12+SSL-1.41 (as well as 1.39, 1.40 and my 1.42)

I install in /usr/local/apache-test/apache-1.3.12+ssl-1.41 for the
purposes of testing. I use the default configuration file with a few
required SSL commands on the end, which also disables SSL (until I
am ready to test it).

I run "httpsdctl configtest" and it says syntax is OK.
I run "httpsdctl start" and it says the following:

  Launching... /usr/local/apache-test/apache-1.3.12+ssl-1.42/bin/gcache
  pid=1284102
  ./httpsdctl start: httpsd started

Looking in the process table, I find that apache and gcache are not
running. Looking in the error_log for the source of failure shows
the following message:

  Syntax error on line 548 of
  /usr/local/apache-test/apache-1.3.12+ssl-1.42/conf/httpd.conf:
  Invalid command 'LogFormat', perhaps mis-spelled or defined by a
  module not included in the server configuration

I get this same error with stock SSL 1.39 through 1.41.

The error seems to point to the fact that the config_log module is not
getting loaded, thus causing the command in the configuration file to
generate a syntax error.

Using alpha-trace on the processes, I find very odd behavior ocurring.
The httpsdctl process fires up a httpsd process. The httpsd process
reads the config file, loads all the apache modules and does various
other things. The observation to note at this point is that httpsd does
indeed process the config file correctly, without errors, and all the
modules are loaded successfully.

Httpsd then forks off a child, which execs into gcache. This happens
without errors. Httpsd then forks off a child, which is itself a httpsd
process (so, now there are 2 httpsd process running). This new httpsd
process then tries to go through the same process that the first one did,
which is read the config file, load the modules and do the things it
needs to do before accepting new connections.

However, the second httpsd process seems to mysteriously skip over lines
in the config file, which means that it misses many of the needed modules,
which happens to include the config_log module.

In the trace file, the usual pattern of opening a module and reading it
into memory is replaced with a bunch of the following:

  close (5) = 0
  syscall_224 () = 0
  socket (AF_UNIX, SOCK_DGRAM, PF_UNSPEC) = 5
  getpid () = 1282755
  sendto (5, 0x11fff9458, 2002, , 2002, 0<1/UNIX, /dev/.dcpid0>,
    1073742512) = -1, Errno 2 (No such file or directory)

This seems to be repeated for the same number of modules that are skipped
in the config file.

Anyways, I seem to be fighting demons here and I just don't understand
what is going on and how to fix it. If anyone has seen this problem, I
would greatly appreciate their help. If anyone has a working current
version of Apache-SSL on their 5.0a system, I would love to hear from
them as well.

I will attach the trace files, as well as the config file, etc. I hope
this will be useful to anyone who wants to help me.

Thanks!
Scott
--
 +-=-=-=-=-=-=-=-=-=+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=+=-=-=-=-=-=-=-=-+
      Scott W. Adkins                http://www.cns.ohiou.edu/~sadkins/
   UNIX Systems Engineer                  mailto:sadkins_at_ohio.edu
        ICQ 7626282                 Work (740)593-9478 Fax (740)593-1944
 +-=-=-=-=-=-=-=-=-=+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=+=-=-=-=-=-=-=-=-+
     CNS, HDL Center, Suite 301, Ohio University, Athens, OH 45701-2979








Received on Fri Aug 04 2000 - 04:09:48 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:41 NZDT