[rabbitmq-discuss] Examining Queue Failover Behavior

[Ben Hood]

Hey Jason,

Don't worry, your post did not disappear down a black hole, we are
just extremely busy doing client related work ATM.

On Mon, Feb 9, 2009 at 5:45 PM, Jason J. W. Williams
<jasonjwwilliams at gmail.com> wrote:
> Hi Ben,
>
> No problem. I appreciate the response. They are durable:
>
>        chan_recv.queue_declare(queue=queue_name,
>                                durable=True,
>                                auto_delete=False)

One thing that I overlooked in your original post was that you
expected a queue process and it's content to be automagically migrated
on the event of a node failure. This does not happen currently. In
general, the process of neatly recovering such that a node failure is
completely transparent to a client is not addressed. The modus
operandi is to a) not lose messages and then b) give somebody the
tools to recover a node. The reason why transparent failover has not
been addressed is because it is quite difficult to do - for example,
how do you when a node has died or that the network has just been
partitioned? Doable, but difficult.

(To Valentino and Co ) as I said before, I'll try to address all of
the related posts today.

Ben