[ejabberd] Clustered deployments

Andy Skelton skeltoac at gmail.com
Fri Jul 24 19:36:10 MSD 2009

Eagerly awaiting your findings. Will the community do a better job of
documenting the software? If so, can we edit the official documentation? The
WordPress Codex comes to mind.


On Jul 24, 2009 11:07 AM, "Carlos Abalde" <carlos.abalde at gmail.com> wrote:

Last e-mail! I promise! :)

> I've looking again into this issue. After some research I've discovered >
the -undocumented?- nod...
I just one to share here a couple of configuration tips related to
component separation I've discovered browsing ejabberd sources.

The undocumented configuration option 'node_type' can have three
possible values:

 - {node_type, generic}. Normal ejabberd behavior.
 - {node_type, frontend}. The node becomes a frontend node. Basically
it handles client connections (parsing included). When a requests
arrives it is delegated in a backend node.
 - {node_type, backend}. The node focus on session management and routing.

Then, you need to define the frontend listening components. In order to
do that, in the listening ports section, you should replace the name of
the module 'whatever' by {frontend, whatever}. For example:

 {5222, {frontend, ejabberd_c2s}, [
                       {access, c2s},
                       {shaper, c2s_shaper},
                       {max_stanza_size, 65536}

Then, I guess, it's necessary to correctly setup distributed mnesia
table replicas. For example, session table should only have RAM replicas
on backend nodes.

I'll continue researching... Any help is welcome! :)


ejabberd mailing list
ejabberd at jabber.ru
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.jabber.ru/pipermail/ejabberd/attachments/20090724/3163a5e5/attachment.html>

More information about the ejabberd mailing list