[rabbitmq-discuss] Mnesia crashes on rmq clustering !

[Tim Watson]

Just to add to what Michael said: no replication scheme in a distributed system will be "issue free", but the right tool(s) will enable you to manage the challenges that distribution entails.

Shovel and Federation both provide a means for replicating state across a wide area (or other) network, but require that you write your client in such a way that it can handle duplication (which is good practise anyway). Clustering/HA provides transparent replication without message loss, but doesn't handle network partitions well.

You can, therefore, decide where to invest effort in building your system - either in preventing or handling (through infrastructure management) partitions, or in handling replication yourself. It's advisable to focus your efforts on (a) the scenario you're most likely to encounter vs the risks involved,  and/or (b) the solution and technology with which you're more comfortable (e.g., managing replication over AMQP or handling network design, dealing with partitions using various sysops tools).

Cheers,
Tim

On 14 Sep 2013, at 10:19, Michael Klishin <michael at rabbitmq.com> wrote:

> tluck:
> 
>> Are shovel and federation are good use here over clustering ? If yes ? then
>> which one is faster and issue free.
> 
> Shovel and federation both use AMQP 0-9-1 connections which will be recovered.
> Note that both are asynchronous replication tools (AP in CAP), while clustering
> is CP.
> 
> See http://www.rabbitmq.com/clustering.html
> and http://www.rabbitmq.com/ha.html.
> 
> MK
> 
> 
> 
> _______________________________________________
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
> https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss