Nope, I spoke too soon. My ExecStart line works for a few seconds and then dies. (noticed that by doing ps ax|grep erl)

But Im sure manually starting works (I been using it this way somedays now)

FYI backslash escapes should work just fine according to the code.

The unit is setup to be Type=forking -- does the service actually fork? I assume this is what the "-detached" flag does. Have you tried removing this flag and starting as Type=simple?

You can get logging from journalctl by looking at the failing PID. That is, grab the PID from systemctl status ejabberd and run 'journalctl -b _PID=<pid>'

No, removing the -detached flag and setting it to simple still does not work.

Here is the output of journalctl:
dic 30 21:52:46 server erl[7199]: {error_logger,{{2012,12,30},{21,52,46}},std_error,"File operation error: eacces. Target: /etc/ejabberd/inetrc. Function: get_file. Process: kernel_sup."}
dic 30 21:52:46 server erl[7199]: {error_logger,{{2012,12,30},{21,52,46}},"inet_config: file ~s not found~n",["/etc/ejabberd/inetrc"]}
dic 30 21:52:46 server erl[7199]: =ERROR REPORT==== 30-Dec-2012::15:52:46 ===
dic 30 21:52:46 server erl[7199]: File operation error: eacces. Target: /etc/ejabberd/inetrc. Function: get_file. Process: kernel_sup.
dic 30 21:52:46 server erl[7199]: =ERROR REPORT==== 30-Dec-2012::15:52:46 ===
dic 30 21:52:46 server erl[7199]: inet_config: file /etc/ejabberd/inetrc not found
dic 30 21:52:46 server erl[7199]: Eshell V5.9.3.1 (abort with ^G)
dic 30 21:52:46 server erl[7199]: (ejabberd@localhost)1> *** Terminating erlang (ejabberd@localhost)

I must add the file /etc/ejabberd/inetrc does exist and has 644 permisions (also tried setting it to 666, but that does not help)

EDIT: Setting the user as root does fix it. Probably there is a permissions problem? I'll continue to do tests. I'll post again as soon as I get any more progress.

EDIT2: Ok, using ejabberd user now. I removed the line for the group on the unit file (the "Group=jabber" part) and then changed the group ownership of the files in the ejabberd home with # chown ejabberd.ejabberd /var/lib/ejabberd/*. It works correctly with this change.
Btw, I tried to conserve the jabber group part and do # chown ejabberd.jabber /var/lib/ejabberd/* but that does not work.

Ok, I think I know what is the problem.
When ejabberd is started by initscripts, the rc file start it by doing:

su ejabberd -s /bin/bash - -c 'erl -pa /usr/lib/ejabberd/ebin -sname ejabberd -s ejabberd -ejabberd config \"/etc/ejabberd/ejabberd.cfg\" log_path \"/var/log/ejabberd/ejabberd.log\" -sasl sasl_error_logger \{file,\"/var/log/ejabberd/sasl.log\"\} -mnesia dir \"/var/lib/ejabberd\" -detached -kernel inetrc \"/etc/ejabberd/inetrc\"'

That starts the service using the ejabberd user, but also, using ejabberd group. (I checked using ps -ae -O fgroup,fuser ). The jabber group is an additional group, is not the main one, and it start the service using the main group of ejabberd user.
I used to use ejabberd on initscripts and worked correctly like that several months.

When the service is started by systemd, is started by using the ejabberd user, BUT using the jabber group this time instead. And it have problems because, at least on my system, /etc/ejabberd has 750 permissions and the ownership is root:ejabberd.

I see two ways to fix it.
The first one is by not using the jabber group at all.
The second one is setting the jabber group as the default group of ejabberd user and then changing the /etc/ejabberd and its files to root:jabber

I did the second one, and is working correctly for me now.

Or simply change the owner of /etc/ejabberd to ejabberd instead of root.

Heh, yeah that works. But, since the maintainer has on the ejabberd.install this line:

chown root:ejabberd /etc/ejabberd/ejabberd.cfg /etc/ejabberd/ejabberdctl.cfg /etc/ejabberd

I guess there is a reason for that... dunno XD

But yeah, changing that to root:jabber or ejabberd.jabber or ejabberd.ejabberd works.

I removed Group= from .service, so systemd should start ejabberd in the same way as initscripts.