[rabbitmq-discuss] cluster & HA (not Mirrored Queue Behaviour)

池凱琳/Kalin Chih kalinchih at gmail.com
Mon Jul 29 17:08:33 BST 2013

 Hi MK,

Thanks for your reply.

For question 1,
my test result is the queue was gone after the rabbitmq1 crashed. And the
rabbitmq2 did not automatically create the queue 'rabbit-test'.
Is it because the producer just send message with exchange & routing key
(no queue name)?

For question 2,
According the rabbitmq cluster guide, it says there is at least one disc
node and ram node is good for performance. I hope my cluster provide good
reliability & performance. But from my testing result of question 1, the
queue eas gone.

So I draft my cluster as below.
- disc node 1 with persistence queues &stats
- disc node 2 mirror with node 1 (replicate the queue?)
- ram node 3 with node 1
- ram node 4 with node 2


Michael Klishin 於 2013年7月29日星期一寫道:

> 2013/7/28 Kalin Chih <kalinchih at gmail.com <javascript:_e({}, 'cvml',
> 'kalinchih at gmail.com');>>
>> Question 1: Is any possible to automatically re-create the queue
>> "rabbit_test" on the other disc node (rabbit at rabbitmq2) when the
>> rabbit at rabbitmq1 is crash?
> Queue metadata (e.g. name, durability, other properties) will be available
> on other nodes automatically.
>> Question 2: What is the best practice for cluster  without "Mirrored
>> Queue Behaviour"?
> It depends on what your goals and needs are. What problem are you trying
> to solve with clustering?
> Why mirroring queue contents is not a good idea in your case?
> --
> MK
> http://github.com/michaelklishin
> http://twitter.com/michaelklishin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20130730/2a8c54bb/attachment.htm>

More information about the rabbitmq-discuss mailing list