[ejabberd] Re: Clustered servers not reconnecting

Matthew Harrell lists-sender-c29da0 at bittwiddlers.com
Sun Dec 10 15:51:30 MSK 2006


> They do that and this is automatic. However, if your two nodes are =20
> both kept online, receiving client connections the two database will =20
> evolves differently. The database are different and thus =20
> inconsistent. Mnesia does not know which version of the data it =20
> should keep, that's why the database wait for a manual operation.
>
> As a first place, I would not build an ejabberd cluster over a wan, =20
> but only over a lan to avoid those problems.
>
> I hope this helps,

Yeah, that helps.  I really just set up the cluster because I thought it
would be useful and a good learning process.  It's not critical in my 
setup.  I may fiddle with reducing the latency or at least determining 
what's causing the high latency in the early morning

When I get the message about the inconsistent, partitioned network I assume
this means that mnesia has detected that they have both run separately for 
a while and can no longer be easily merged.  Is there a process I can use 
at this point to get them back together?

-- 
  Matthew Harrell                          Quantum Mechanics:
  Bit Twiddlers, Inc.                       The dreams stuff is made of.
  mharrell at bittwiddlers.com     


More information about the ejabberd mailing list