[OPEN-ILS-DEV] ejabberd problems

[Dan Scott]

Hi Cyrus:

2008/7/14 Lyons, Cyrus 2 <lyonsc2 at lsbu.ac.uk>:
> Dear all,
>
> Apologies in advance if this has already been looked at, but extensive
> searching hasn't yielded many results.
>
> I'm having a number of issues after an automated install on Debian
> Etch. This is actually my second install on this machine: The first
> had similar problems which I assumed were due to numerous hostname
> changes throughout it's life, and I figured it would be quicker to
> just build the computer cleanly to get rid of any old config files and
> start from scratch than to weed out the problem. Bad call,
> unfortunately - having identical problems now, so perhaps a waste of
> time...

Perhaps. You said you did an automated install - was it using John
Schmidt's script? This might help us narrow down the problem. What
version of Evergreen?

In the Debian Etch install instructions, I think I'll insert a "If you
run into problems with the "login via srfsh" step, stop now. Run
Open-ILS/src/support-scripts/settings-tester.pl; if that doesn't
provide any helpful feedback, post your problem to the open-ils-dev
mailing list before making any drastic changes to your system
configuration. You're probably closer to a solution than you think."
(something along the lines of the advice to sit still rather than keep
wandering in a forest when you're lost).

> The issues are as follows.
>
> First, after the install, running the admin login command failed:
>        root at evergreen:~# /openils/bin/srfsh
>        srfsh# login admin open-ils
>        ???: login admin open-ils
>        srfsh# exit
>
> I had seen various error messages thrown during the install script
> itself, but unfortunately flashing past too quickly to really take
> notes.
>
> ~/srfsh.log yielded some information:
>        [/*--S
> osrfMessage--*/{"threadTrace":"0","type":"CONNECT"}/*--E
> osrfMessage--*/]
>
>        srfsh 2008-07-14 07:42:44 [DEBG:9022:osrf_stack.c:73:] We
> received 1 messages from router at evergreen.lsbu.ac.uk/open-ils.auth
>        srfsh 2008-07-14 07:42:44 [WARN:9022:osrf_stack.c:84:]  !!!
> Received Jabber layer error message
>        srfsh 2008-07-14 07:42:44 [WARN:9022:osrf_stack.c:94:]  *
> Jabber Error is for top level remote id
> [router at evergreen.lsbu.ac.uk/open-ils.auth], no one to send my message
> too!!!
>        srfsh 2008-07-14 07:42:44 [INFO:9022:osrf_stack.c:103:]
> Message processing duration 0.000065
>        srfsh 2008-07-14 07:42:44 [DEBG:9022:osrf_stack.c:106:] after
> msg delete
>        srfsh 2008-07-14 07:42:44 [DEBG:9022:osrf_app_session.c:554:]
> AppSession in queue_wait with timeout 5
>        srfsh 2008-07-14 07:42:49 [DEBG:9022:osrf_app_session.c:554:]
> AppSession in queue_wait with timeout 0
>        srfsh 2008-07-14 07:42:49 [WARN:9022:srfsh.c:539:] Unable to
> connect to remote service open-ils.auth

How did you start Evergreen - did you use start_all or start_router /
start_perl / start_c ? Just asking because the former method can
invoke a race condition that prevents the C services from starting -
and open-ils.auth is a C service.

> I read somewhere that it'd be a good idea to clean up any reference to
> 'localhost' in ejabberd.cfg, openils.xml, and openils_core.xml. After

Actually, the current thinking is that that's a really bad idea on a
single-server machine. The only hostname that needs to be changed on a
single-server machine is the hostname in the "<hosts>" entry in
opensrf.xml - and even that requirement will be removed in Evergreen
1.4. Changing hostnames in ejabberd.cfg can really upset ejabberd.

> doing so, went to stop apache2, openils, and ejabberd (and then start
> in reverse order), but now having problems restarting (assuming they
> were really running before, which I now doubt) opensrf and apache:
>        opensrf at evergreen:~$ osrf_ctl.sh -c
> /openils/conf/opensrf_core.xml -a start_router
>        Starting OpenSRF Router
>        !!!! Unable to connect router to jabber server
> evergreen.lsbu.ac.uk... exiting
>        opensrf at evergreen:~$ exit
>        logout
>        root at evergreen:/openils/conf# /etc/init.d/ejabberd start
>        Starting jabber server: ejabberd already running.
>
> Didn't quite see the point of trying perl and c, but here goes
> anyway...
>        opensrf at evergreen:~$ osrf_ctl.sh -c
> /openils/conf/opensrf_core.xml -a start_perl
>        Starting OpenSRF Perl
>        opensrf at evergreen:~$ osrf_ctl.sh -c
> /openils/conf/opensrf_core.xml -a start_c
>        Starting OpenSRF C (host=evergreen.lsbu.ac.uk)
>        Ctrl + C
>
> As you can see, Perl starts fine; C hangs around seemingly doing
> nothing and needs me to Ctrl+C to be able to carry on. Next up,
> apache:
>        root at evergreen:/openils/conf# /etc/init.d/apache2 start
>        Starting web server (apache2)...Syntax error on line 51 of
> /etc/apache2/sites-     available/eg_vhost.conf:
>        Invalid command 'XMLEntStripPI', perhaps misspelled or defined
> by a module not included in the server configuration
>         failed!

Hmm, your apache server should be loading a new apache module named
mod_xmlent.so (which will give it the XMLEntStripPI command, among
others) and osrf_json_gateway.so as part of /etc/apache2/httpd.conf.
Perhaps apache was already running?

> I will attach opensrf.xml and opensrf_core.xml if necessary, but this
> message seems long enough without, and I thought someone might have
> some inspired counsel to offer.

First thing I would try doing is running settings-tester.pl - if that
turns up nothing obvious, then I would work towards getting
ejabberd.cfg back to recognizing localhost. You might need to purge
the package entirely with dpkg to remove all traces of it, then
reinstall ejabberd and follow the ejabberd configuration steps in
http://open-ils.org/dokuwiki/doku.php?id=installing_evergreen_1.2_on_debian_etch_x86_32-bit
starting at step 12.

Keep us informed!

-- 
Dan Scott
Laurentian University