[ejabberd] S2S overloaded

Carlos Abalde carlos.abalde at gmail.com
Tue Aug 25 19:18:04 MSD 2009


Hi all,

Today I've testing a clustered deployment of ejabberd, each cluster
running only one jabber domain. Rosters of users are composed by members
of almost every domain, therefore there is a lot of S2S communication.

After some stress testing (just connections/disconnections with the
associated presences) I reached a limit: ejabberd_s2s_in processes
started queuing requests in their mailboxes (up to message_queue_len's
greater than 400000 stanzas in one of the nodes!).

After reviewing implementation of S2S communication in ejabberd I'm not
able to find the bottleneck. In fact the ejabberd_s2s_in behavior is
pretty simple. Any ideas/help on how to improve the S2S throughput? Is
there any way of starting several ejabberd_s2s_in processors balancing
incoming S2S requests among them?

Thanks!

-- 
Carlos.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: OpenPGP digital signature
URL: <http://lists.jabber.ru/pipermail/ejabberd/attachments/20090825/92331d2f/attachment.pgp>


More information about the ejabberd mailing list