XMailforum is a readonly knowledge archive now.

Registering as a new user or answering posts is not possible anymore.

Might the force be with you, to find here what you are looking for.

2019-09-20 - hschneider, Admin

Cookie Disclaimer: This forum uses only essential, anonymous session cookies (xmailforum*), nothing to be scared of.

XMail Forum [Powered by Invision Power Board]
Printable Version of Topic
Click here to view this topic in its original format
XMail Forum > XMail Queue Manager > XQM Agent Installation Guide


Posted by: hschneider Jul 10 2003, 11:11 AM
QUOTE

How to install:

  - Login as root
  - Unpack the archive and copy the Agent's directory to your preferred location.
  - Do NOT move any of the *.so files to another location.
  - Modify xqmagent.conf to your needs.
  - Move  xqmagent.conf to /etc/xqmagent.conf
  - Start xqmagent directly or in one of your system startup scripts
 
  The Agent immidiately detaches from the console and runs as a daemon process
  transparently in background.

How to debug:

  ./xqmagent --debug
 
  This shows you the Agen't configuration and lets you monitor the
  remote procedure calls being processed.

  Debugging is generally a good idea to verify that all your settings are
  valid and that the Agent is bound correctly to its port.

  Debugging requires Agent >= v1.24


Troubleshooting ...

  How to overcome some nuts and bolts during the Agent's startup:
 
    "ERROR: Cannot access /etc/xqmagent.conf ..."
    Make sure, the .conf file is placed in your /etc folder.
 
    "ERROR: Environment variable MAIL_ROOT not found! ..."
    Locate your xmail startup script. E.g. on a SUSE Linux distribution
    its path is /etc/rc.d/init.d/xmail
   
    Make sure that there is a statement
    export MAIL_ROOT
    in the head of the script or inside the start_xmail() function.
   
    If this is OK, then wrap the agent's call into a startup script,
    which contains
   
    #!/bin/sh
    export MAIL_ROOT=/var/MailRoot
    ./xqmagentwin
   
    Don't forget to chmod 755 scriptname ...
   
    However the most easy way is to force a mailroot folder in the
    .conf file (this requires Agent >= 1.24) ;-)

  On some Linux systems: 
   "ERROR:
   File "<string>", line 3, in ?
   File "iu.py", line 274, in importHook
   [...]
   ImportError: libssl.so.0.9.6: cannot open shared object [...]"

  
   Ouch .. this is a Python backtrace, which tells you that libssl.so.0.9.6
   cannot be found. This lib is part of the openssl package which should be
   included in your Linux or FreeBSD distribution.
   
   At first, try to locate libssl.so.0.9.6 on your system.
    
   If you cannot find it:
   - Install the openssl package from your package manager
   - If the lib is NOT installed to /usr/lib or
     if the lib is NOT exactly named libssl.so.0.9.6 (it might end with .6b
     or something else) then do
      
     cd /usr/lib
     ln -s location_of_libssl/libssl.so.0.9.6b libssl.so.0.9.6

     This creates a symbolic link wich pretends to be the correct version.
      
   The lib is installed but the error is still there:
   - Make sure that /usr/lib is in your search path
   - Some package managers place the lib in /lib and not in /usr/lib in that
     case you should
    
     cd /usr/lib
     ln -s /lib/libssl.so.0.9.6b libssl.so.0.9.6

  "ERROR: ... address already in use!"
    - You have assigned a port that is already used by another application.
      Change AGENT_PORT in the .conf file.
    - You have entered a non-existing AGENT_IP in the .conf file.
      Remember to put the Agent's machine IP here, NOT XQM's IP.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

  If you stumble over some other error, don't hestitate to contact me.
  Mail the Agent's output to h_schneider@marketmix.com
  Include the Agent's version and your platform as well.
 
  This software can only improved with your feedback!


-- Harald


Posted by: hschneider Sep 6 2003, 06:46 AM
System:
Redhat 9.0

Symptom:
Openssl is installed and /usr/lib is in my PATH, but I still get
ImportError: libcrypto.so.0.9.6: cannot open shared object [...]"

Solution:
Create a symbolic link from version 0.9.6 to your current lib version, e.g.
ln -s /lib/libcrypto.so.0.9.7a /lib/libcrypto.so.0.9.6

Don't confuse /lib with /usr/lib here. RedHar searches /lib at first!

Posted by: hschneider Sep 18 2003, 11:54 AM
System:
All Linux platforms

Symptom:
Trying to start ./xqmagent results in a Python backtrace dump:

QUOTE

mod is NULL - archiveTraceback (most recent call last):
File "archive.py", line 18, in ?
ImportError: No module named struct
Traceback (most recent call last):
File "<string>", line 2, in ?
AttributeError: 'archive' module has no attribute 'PYZOwner'
Traceback (most recent call last):
File "<string>", line 1, in ?
ImportError: No module named codecs
Traceback (most recent call last):
File "<string>", line 3, in ?
ImportError: No module named daemonlib


Solution:
The Agent does not find its additional files coming with the original archive.

Make sure that all *.so files reside in the same folder like xqmagent and that their permissions are set properly. Do NOT copy any of its .so files to another location.

Posted by: hschneider Sep 26 2003, 12:20 PM
System:
Some Linux platforms, e.g. RedHat 9.

Symptom:
The Agent cannot restart XMail server, even after defrosting.

Solution:
Upgrade to Agent >= 1.30
Check your your /etc/xqmagent.conf, if STARTSCRIPT holds a valid path to your XMail startscript.

Since Agent 1.30 starting and restarting the server is done the save way via XMail's startscript. This requires the full path to your startscript in /etc/xqmagent.conf. E.g. STARTSCRIPT=/etc/rc.d/init.d/xmail

Posted by: hschneider Nov 14 2003, 09:59 AM
System:
RedHat 9.0

Symptom:
XQM hangs, when querying the queue with the Agent.

Solution:
The way how threads are handled by the kernel has changed since RedHat 9.
See http://www.redhat.com/docs/manuals/linux/R...ease-notes/x86/ for details.

Since XQM and XQMAgent use threads extensively, they are both affected by this change.

For better compatibility, you should wrap them into a startup script which forces the system to emulate the old thread scheme:

------------------------------
#!/bin/sh
export LD_ASSUME_KERNEL=2.4.1
./xqmagent # or ./xqmwin
------------------------------

Don't forget to do
chmod 755 scriptname

Posted by: hschneider Nov 24 2003, 07:44 AM
System:
Linux

Symptom:
The Agent does not restart XMail server after a 'Stop Server' or 'Defrost'.

Solution:
You use an older /etc/xqmagent.conf file. Just add

STARTSCRIPT=/etc/init.d/xmail

(or whatever your startscript is) to the .conf file.

Posted by: hschneider Nov 25 2003, 08:37 AM
System:
Tawie Linux

Symptom:
Openssl is installed, /usr/lib is in my PATH and I created soft links for libssl and libcrypto, but I still get
ImportError: libcrypto.so.0.9.6: cannot open shared object [...]"

Solution:
Tawie seems not to accept soft links to access libraries. Therefore do some real copies. E.g. if you have installed openssl 0.9.7:

cd /usr/lib
cp libssl.so.0.9.7 libssl.so.0.9.6
cp libcrypto.so.0.9.7 libcrypto.so.0.9.6

Posted by: hschneider Jan 22 2004, 02:42 PM
System:
Solaris 8 and lower.

Symptom:
Starting ./xqmagent results in

QUOTE

Traceback (most recent call last):
File "xqmagent.py", line 4, in ?
ImportError: No module named daemonlib


Solution:
Install the postinstall fix from http://xmail.marketmix.com

Posted by: hschneider Apr 20 2004, 10:32 AM
System:
FreeBSD 5.0 and higher

Symptom:
Starting ./xqmagent results in

QUOTE

/usr/libexec/ld-elf.so.1: Shared object "libc_r.so.4" not found


Solution:
You have to install the FreeBSD 4.6 compatibility layer:
CD1/misc/compat4x-XXX-5.0.XXXXXX

Posted by: hschneider Apr 20 2004, 10:54 AM
System:
FreeBSD 5.0 and higher

Symptom:
Starting ./xqmagent results in

QUOTE

/usr/libexec/ld-elf.so.1: Shared object "libutil.so.3" not found


Solution:
This seems to be a bug on the FreeBSD 5 development branch.
Here its fix:

cd /lib
ln -s libutil.so.4 libutil.so.3

Posted by: hschneider Jul 8 2004, 09:19 AM
System:
NetBSD

Symptom:
XQM dies with a coredump after being contacted by XQM.
XQM says 'Agent not reachable'.

Solution:
Start the Agent this way:
./xqmagent --debug >/dev/null &

This avoids it's process to be detached the usual way, which causes problems with Python's threading engine on this platform.

Posted by: hschneider Jan 26 2005, 06:36 PM
System:
Debian (and maybe other Unix flavours)

Symptom:
Application startup fails with:
socket.gaierror (-2; 'Name or service not known')

Solution:
This error is related to DNS resolution.
It comes directly from Python's socket support and is caused by a failed resolution on OS level. Use fixed IP addresses in XQM's (or the Agent's) config.
Additionally you should check your DNS configuration.

Posted by: hschneider Dec 10 2005, 08:42 AM
System:
All Linux and BSD systems with glibc < 2.2

Symptom:
QUOTE

I start xqmagent and i receive the message
./xqmagent: /lib/libc.so.6: version `GLIBC_2.2' not found (required by ./xqmagent)
I have glibc 2.1.3. An upgrade is impossible


Use XQMAgent/generic which runs on any system with Python 2.3x installed.

Download the file attached, unpack and start with
python xqmagent.pyc
or
python xqmagent.pyc --debug

Python can be found here: http://www.python.org
Use Version 2.3.x, not 2.4.x if possible.

Powered by Invision Power Board (http://www.invisionboard.com)
© Invision Power Services (http://www.invisionpower.com)