[ejabberd] Cluster Registration and Use

Bill Brazell bill.brazell at cingular.com
Tue May 3 18:13:17 MSD 2005

I'm trying to get a two-node ejabberd cluster working with a hardware load
balancer and am having trouble.  First, let me explain the setup...

Server1 - ejabberd node 1 (ejabberd at Server1.longname.com)
Server2 - ejabberd node 2 (ejabberd at Server2.longname.com)
LeadName - name of the front end of the load balancer address for the ejabberd
pool (LeadName.longname.com)  The ejabberd pool is set to use persistent

Server1 and Server2 are recognized as running nodes within the web interface.
An admin account (admin1 at Server1.longname.com) was created successfully on
Server1.  Access to the web interface using the LeadName and the account
credentials for admin1 (admin1 at Leadname.longname.com) is granted successfully.

When I attempt to create a new account such as user1 at LeadName.longname.com, it
fails.  I can see the packets hit the interface of Server1, but the account
creation fails.  Account registration works properly if I bypass the LeadName
and go straight to Server1.

When I attempt to register and account in the clear over port 5222, this entry
is created in the ejabberd.log:

=INFO REPORT==== 2005-05-03 10:00:28 ===
I(<0.909.0>:ejabberd_listener:90): (#Port<0.588>) Accepted connection
{{10,87,201,196},1167} -> {{10,87,202,152},5222}

When I attempt to register and account via SSL over port 5223, this entry is
created in the ejabberd.log:

=INFO REPORT==== 2005-05-03 10:04:22 ===
I(<0.910.0>:ejabberd_listener:142): ({sslsocket,5,<0.6126.12>}) Accepted SSL
connection {{10,87,201,196},1169} -> {{10,87,202,152},5223}

=ERROR REPORT==== 2005-05-03 10:04:23 ===
E(<0.6128.12>:ejabberd_receiver:88): ({sslsocket,5,<0.6126.12>}) abnormal
ssl:recv termination:

I also see repeated entries in the log such as this when no client connection
attempts are made:

=INFO REPORT==== 2005-05-03 10:05:56 ===
I(<0.910.0>:ejabberd_listener:153): ({sslsocket,4,<0.915.0>}) Failed SSL
handshake: esslaccept

I did create new certificates for both nodes using LeadName.longname.com as the
Common Name.  SSL problems aside, any idea why the port 5222 connection was
accepted but not properly processed?

I must have some major configuration problems going on here.  Anyone have any
ideas how to get the connections to work through the load balancer?

Thanks in advance!

More information about the ejabberd mailing list