[ejabberd] ejabberd clustering/dialback issue

Vasil Kolev vasil at ludost.net
Fri Aug 26 17:29:24 MSD 2011


Hello,

I have a simple ejabberd cluster, on two machines in two separate
locations. Here are the SVR records:

_xmpp-client._tcp 900 IN SRV 5 0 5222 marla.ludost.net.
_xmpp-client._tcp 900 IN SRV 5 0 5222 tyler.ludost.net.

_xmpp-server._tcp 900 IN SRV 5 0 5269 marla.ludost.net.
_xmpp-server._tcp 900 IN SRV 5 0 5269 tyler.ludost.net.

_jabber._tcp      900 IN SRV 5 0 5269 marla.ludost.net.
_jabber._tcp      900 IN SRV 5 0 5269 tyler.ludost.net.


The problem is that after a while I start to get the following error:
I(<0.17861.1>:ejabberd_s2s_out:353) : Closing s2s connection: ludost.net -> jabber.org (invalid dialback key)
I(<0.439.0>:ejabberd_s2s_out:353) : Closing s2s connection: ludost.net -> nikolay.bg (invalid dialback key)

This happens with all kinds of destinations, a restart of both ejabberd
daemons fixes it.

The connection between the two machines is usually ok, but not always
(there are some drops in it,  here's a smokeping for that link
http://ludost.net/cgi-bin/smokeping.cgi?target=Connectivity.tyler ), is
that the reason for this issue, and is there any way to survive those
issues?

Also, is there a solution for when one of the machines goes away (e.g.
if the datacenter dies) for the connections to the other servers to
work, as usually some of them try to connect to the wrong one and
timeout? E.g. is there a way, for example, to add a 'connect timeout' in
the SRV record ?:)


-- 
Regards,
Vasil Kolev
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <http://lists.jabber.ru/pipermail/ejabberd/attachments/20110826/ea6688bc/attachment.pgp>


More information about the ejabberd mailing list