[ejabberd] Issue with S2S

Julien Genestoux julien.genestoux at gmail.com
Sun Aug 7 02:47:13 MSD 2011


Stephan,
yes, I think your schema sums up things correctly. One idea that comes to
mind (and that is not represented in your schema) is that there are 2
service.com and many component.service.com. Of course, the 2 service.com are
in a cluster, and they are both in the SRV records for service.com... and
the domain component.service.com points via an A record to both
service.comservers, with the same priority. Could that be an/the
issue?

Also, I am sure that the components are not faulty, because IQs are always
received by anyuser at service.com if it makes similar requests.
The issue lies after service.com, or on consumer.com, and yes the from/to
addresses are working just fine.  I am not sure I understand " Could you
find out if the
responses reach the s2s boundary of systems at equal probability?" Do you
mean that I should make sure that both XMPP servers for service.com get
stanzas from component.service.com? I have done tests connecting to only one
of these with a local client and they both work fine/as expected...

Julien



On Sat, Aug 6, 2011 at 11:39 PM, Stephan Maka <stephan at spaceboyz.net> wrote:

> Julien Genestoux wrote:
> > On the other hand, we have another consumer.com domain, which hosts 1
> > ejabberd server (at consumer.com), and on this server, we have local
> > clients. These clients need to interract with the components over at
> > service.com, and that's where the issues start.
> > Basiclaly, client at consumer.com sends IQs, and wait for responses. The IQ
> are
> > received by the components at component.service.com, and they respond
> just
> > fine... but the responses do not reach client at consumer.com reliably.
> > Sometimes they do, sometimes they don't. I noticed that they tend to
> arrive
> > more when the consumer.com server has been restarted recently, and they
> tend
> > to never arrive when the server has been up for a couple hours. I am
> trying
> > to find out why this is happening.
>
> Summarizing your setup:
>
> +--------------+            +-------------+
> | consumer.com |------------| service.com |
> +--------------+            +-------------+
>      |                            |
> +-----------------+ +---------------------+
> |user at consumer.com| |component.service.com|
> +-----------------+ +---------------------+
>
> The problem: user at consumer.com can send <iq/> requests to
> component.service.com, but the responses sometimes get lost. Right?
>
> Are the stanza's from/to addresses correct? Could you find out if the
> responses reach the s2s boundary of systems at equal probability?
>
>
> Kudos,
> Stephan
>
> PS: Don't forget Chaos Communication Camp 2011, starting on wednesday.
>
> _______________________________________________
> ejabberd mailing list
> ejabberd at jabber.ru
> http://lists.jabber.ru/mailman/listinfo/ejabberd
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.jabber.ru/pipermail/ejabberd/attachments/20110807/fcc7916f/attachment.html>


More information about the ejabberd mailing list