[rabbitmq-discuss] Shovel failover
tim at rabbitmq.com
Thu May 2 16:02:46 BST 2013
On 2 May 2013, at 15:43, Laing, Michael P. wrote:
> Is the new failover scheme described somewhere? Couldn't find it.
It's described briefly in the "Getting Started" section of http://www.rabbitmq.com/shovel.html.
> Some of my configurations rely on shovels being set up differently for
> each node in a cluster and that the shovel will stick to the node on which
> it is configured.
I'd suggest taking a look at what it does now then. If the same (named) shovel is defined on two nodes that are clustered, it will run on only one of them.
> The reason for these configurations (backup strategy to compensate for
> partitioning) may have become moot nowŠ
Perhaps - please let us know if that's not the case.
> However, I have gotten pretty used to the shovel just being a
> straightforward, well-behaved application on a single node, driven by that
> node's configuration. If it now jumps around among nodes in the cluster
> automatically then it is no longer straightforward - let's hope it is
> still well-behaved!
Indeed. Sadly, it looks like we have broken the default reconnect_delay setting that was previously available. I've filed a bug to fix that. Thanks for bringing it to our attention.
More information about the rabbitmq-discuss