Why Does Ejabberd Start Fail?

Question

I am trying to install ejabberd 2.1.10-2 on my Ubuntu 12.04.1 server. This is a fresh install, and ejabberd is never successfully installed.

The Install

Every time, apt-get hangs on this:

Setting up ejabberd (2.1.10-2ubuntu1) ...
Generating SSL certificate /etc/ejabberd/ejabberd.pem...

Creating config file /etc/ejabberd/ejabberd.cfg with new version
Starting jabber server: ejabberd............................................................ failed.

The dots just go forever until it times out or I 'killall' beam, beam.smp, epmd, and ejabberd processes. I've turned off all firewall restrictions.

Here's the output of epmd -names while the install is hung:

epmd: up and running on port 4369 with data:
name ejabberdctl at port 42108
name ejabberd at port 39621

And after it fails:

epmd: up and running on port 4369 with data:
name ejabberd at port 39621

At the same time (during and after), the output of both netstat -atnp | grep 5222 and netstat -atnp | grep 5280 is empty.

---

The Crash File

A crash dump file is create at /var/log/ejabber/erl_crash.dump. The slogan (i.e. reason for the crash) is:

Slogan: Kernel pid terminated (application_controller) ({application_start_failure,kernel,{shutdown,{kernel,start,[normal,[]]}}})

---

It's alive?

Whenever I try to relaunch ejabberd with service ejabberd start, the same thing happens - even if I've killed all processes before doing so.

However, when I killall the processes listed above again, and run su - ejabberd -c /usr/sbin/ejabberd, this is the output I get:

Erlang R14B04 (erts-5.8.5) [source] [64-bit] [rq:1] [async-threads:0] [kernel-poll:false]

Eshell V5.8.5  (abort with ^G)
(ejabberd@ns1)1> 
=INFO REPORT==== 15-Oct-2012::12:26:13 ===
I(<0.478.0>:ejabberd_listener:166) : Reusing listening port for 5222

=INFO REPORT==== 15-Oct-2012::12:26:13 ===
I(<0.479.0>:ejabberd_listener:166) : Reusing listening port for 5269

=INFO REPORT==== 15-Oct-2012::12:26:13 ===
I(<0.480.0>:ejabberd_listener:166) : Reusing listening port for 5280

=INFO REPORT==== 15-Oct-2012::12:26:13 ===
I(<0.40.0>:ejabberd_app:72) : ejabberd 2.1.10 is started in the node ejabberd@ns1

Then, the server appears to be running. I get a login prompt when I access http://mydomain.com:5280/admin/. Of course I can't login unless I create an account.

At this time, the output of netstat -atnp | grep 5222 and netstat -atnp | grep 5280 is as follows:

tcp        0      0 0.0.0.0:5222            0.0.0.0:*               LISTEN      19347/beam      
tcp        0      0 0.0.0.0:5280            0.0.0.0:*               LISTEN      19347/beam      

---

ejabberdctl

Even when it appears ejabberd is running, trying to do anything with ejabberdctl fails. For example: trying to register a user:

root@ns1:~# ejabberdctl register myusername mydomain.com mypassword
Failed RPC connection to the node ejabberd@ns1: nodedown

---

I have no idea what I'm doing wrong. This happens on two different servers I have with identical software installed (really not much of anything). Please help. Thanks.

Solution

I had this problem one time when the daemon could not determine the ip address of the hostname I gave in the ejabber config file. My solution was to edit /etc/hosts file and make sure that there was a hostname entry for my public ip address and the domain I wanted ejabber to respond to.

0.0.0.0 hostname.domain.com hostname
1.1.1.1 ns1.yahoo.com ns1

Well, you get the picture. Hope that helps.

Comments

[Admin]

Just a guess, have you tried either acquiring a binary/package from the developer web site or building from source? Also, have you tested installing it in a VM (or actual install) of 12.10?

[Andrew Ensley]

I checked, and I already have an entry in my /etc/hosts file with my domain name and public IP. However, I think you're on the right track with this. I just noticed the error message says "Failed RPC connection to the node ejabberd@ns1". ns1 is the machine's name but it is not the domain name. Should that show the domain name there?

[StuckAt7]

yeah for sure.. You should at least add ns1 to the hosts file as well. like 0.0.0.0 <FQDN> <hostname>

[Andrew Ensley]

Holy... I don't even... Adding ns1 to my /etc/hosts file worked. Good grief.

[StuckAt7]

Glad to hear you got it taken care of. Good luck and may the force be with you. :-)