[rabbitmq-discuss] Clustering over a WAN

[Simon MacMullen]

On 28/02/11 16:17, Chris Hampson wrote:
> Thanks Simon and Matthew for the quick responses.
>
> I was beginning to figure that perhaps what we were trying to do was
> going to be problematic, which is a shame as it would've been a very
> neat solution to what we're trying to do.
>
> In essence our use case is as follows:

<snip>

> This was essentially it, we then planned other applications on top of
> this base, it was shaping up nicely until we started experiencing the
> cluster break ups :-\
>
> I've just looked into the shovel plugin further, and it looks as if
> it may not be quite dynamic enough for our use (we don't want to be
> re-tweaking config files every time we set up a new pattern) :-( it's
> beginning to look like the best way forward for now is separate
> brokers with custom clients acting as the shovels between sites,
> consuming on one, then publishing to another as necessary.

You may be right for the time being. Which is a pity as yours is exactly 
the problem the new federation plugin is intended to solve, but it's not 
ready yet.

There's some documentation at:

http://hg.rabbitmq.com/rabbitmq-federation/file/default/README

If it sounds interesting then you might want to play with it, although I 
warn you that it's still under heavy development. But if you can live 
with a temporary solution for a while I intend for it to be more 
generally useful soon.

Cheers, Simon

-- 
Simon MacMullen
Staff Engineer, RabbitMQ
SpringSource, a division of VMware