Matias | 28 Jul 14:08
Picon

Re: Cluster problem with mod_shared_roster

El 28/07/10 13:51, Badlop escribió:
> 2010/7/28 Matias<matiassurdi <at> gmail.com>:
>> I'm running ejabberd 2.1.4 in a two node cluster. They run perfectly, I can
>> connect to a node and chat with another user on the other node, everything
>> works perfect.
>>
>> The problem arises when I setup shared groups with mod_shared_roster: After
>> doing that, on one of the nodes everything works fine, but users on the
>> second node isn't getting the buddy list.
>
> I imagine your problem is this:
>
> mod_shared_roster adds groups to the roster stanzas returned to clients,
> at runtime. It doesn't modify the roster stored in the database.
>
> This means that you must enable this module on both nodes.
>
> Also, you must ensure that the module has similar shared roster groups
> in both nodes:
> A) Either recreate the groups in your second node, using again WebAdmin
> B) Or configure the mnesia tables sr_user and sr_group to be
> synchronized between both nodes.
>
>
> ---
> Badlop
> ProcessOne

Thanks for your quick reply:

The configuration file is exactly the same on both nodes, the only 
module that is not on the second is mod_ctlextra.

I created the shared groups with ejabberdctl srg-* commands, on 
ejabberd <at> first. When I connect to the node ejabberd <at> second, evrything 
works fine.

The problem arises when I try to connect to ejabberd <at> first (the one from 
which I created the groups).

Regarding to your point B, How can I ensure that the tables are in sync? 
(I have set them to keep RAM and disk copies on both nodes).

Thanks for your help.

Gmane