s2s reconnection after a down node.

Hello,

I have Jabber domain with 2 ejabberd nodes (on 2 different machines). This setup works fine. Able to login to one and talk to users on the other and use MUC etc.

However if node A goes down (init:stop()) and comes back online clients on node B are unable to use services defined on A (notably MUC).

Also if clients on B logoff and attempt a login back to B I get the following error:

=ERROR REPORT==== 8-Mar-2005::11:52:48 ===
Discarding message (3,1){route,{jid,"totem","jabber.kataragama.com",[],"totem",
"jabber.kataragama.com",[]},{jid,"totem","jabber.kataragama.com",
"tkabber","totem","jabber.kataragama.com","tkabber"},
{xmlelement,"iq",[{"id","2"},{"type","error"}],
[{xmlelement,"query",[{"xmlns","jabber:iq:privacy"}],
[{xmlelement,"active",[{"name","i-am-visible-list"}],[]}]},
{xmlelement,"error",[{"code","404"},{"type","cancel"}],
[{xmlelement,"item-not-found",
[{"xmlns","urn:ietf:params:xml:ns:xmpp-stanzas"}],[]}]}]}}
to a process in an old incarnation of this node

The client goes into a trying-to-login-and-waiting-for-server state. Browsing to either A or B using the Web interface shows
the client authenticated of both A and B.

Basically my question is: given a clusted ejabberd deployment
how would you reestablish s2s communication when one of the nodes
goes down.

Thanks

totem