[rabbitmq-discuss] routing threads on a rabbitmq node

[Matthew Sackman]

On Thu, Feb 04, 2010 at 10:23:26PM -0800, Brian Sullivan wrote:
> so in the shovel model, what happens if one of the downstream
> topic-based nodes crashes? Seems like all consumers on that node would
> lose messages until we shut down the producing side, correct?

If you use named queues on the central node, there is nothing to stop
the leaf node reconnecting to it and draining it when it comes back up.
Obviously if you use anonymous queues then that gets somewhat harder.
Thus again, you'd have to manage some naming scheme yourself.

> The
> volume is likely too high to buffer in memory even if the shovel
> queues were able to survive until reconnect.

Hmm, you're very welcome to try the new persister, branch bug21673,
which will be in the next major release of RabbitMQ. The branch
certainly will be able to manage memory effectively and push messages to
disk. Unfortunately, until that branch is QA'd and is released, I'm not
sure what else to suggest.

Matthew