[rabbitmq-discuss] Disconnecting vs failure of a disk node and reconnecting a new one...?

Navigateur naveen.chwl at gmail.com
Fri Jul 6 22:13:32 BST 2012

Hi Francesco, thanks for your answer (howcome it isn't showing up in this 
google group)? Can you give me an example of a durable resource 
creation/deletion operation that would benefit from a RAM node (so that I 
know what that means)? If I do have a lot of these, then would I benefit 
from using the 3rd (non-wep-app) machine as a RAM node, along with the 2 
(relevant web app containing) disc nodes? Also, how do I approximately 
calculate the maximum disk space the disc nodes may possibly need in the 
very worst case scenario, say when there are 1 Mb/s of messages?

On 6 July 2012 18:25, Francesco Mazzoli wrote:
> Hi Navigateur,
> At Thu, 5 Jul 2012 05:57:57 -0700 (PDT),
> Navigateur  wrote:
>> New to RabbitMQ. I want do to high-availability over 2 machines. Do I 
>> them both disk nodes or just 1?
> Having only one disc node in a cluster deprives you of the nicest 
> of a cluster: you can be sure that if one database state gets corrupted 
for some
> reason, you still have the other one. This is even more relevant when 
using HA
> queues, which is what you want to do.
> Having only a RAM node online is a situation that you should avoid - 
losing data
> is really easy in that position.
>> If just 1, but its disk fails, does the other one automatically become a 
>> node"?
> No.
>> If not, and I replace the failed-disk machine with a brand new machine 
>> the same name (e.g. rabbit at server1) and simply instruct it to join the 
>> - will everything continue to work as if nothing happened?
> Yes, that should work.
>> If I can't, then 2 disk nodes is the better solution, yes? And if I did 
>> is recovering from an abrupt failure of 1 of the machines simply to 
>> that machine, give it the same name (e.g. rabbit at server1) and call the
>> rabbitmqctl cluster command? Is that all I would have to do?
> Again, that should be OK.
>> I also have a 3rd machine running, which doesn't have any local web app
>> software which creates or listens for any RabbitMQ messages. Would I 
>> from using that as an additional (3rd) RabbitMQ node still? For example, 
>> 3rd non-web-app machine as a disk node, and the 2 web-app-clustered 
>> load-balance) machines as RAM nodes? And maybe have a RAID disk in the 
>> non-web-app machine. Is this the best solution of all? Last question, 
how do I
>> recover after explicitly removing a disk node (i.e. not through disk 
>> this in addition do the disk-failure question)?
> See my reply at the top of this message. In general you'd want to measure 
if RAM
> nodes give you a real benefit and consider them once you established 
that. RAM
> nodes do (durable) resource creation/deletion faster, so if you are using 
> operation intensively you might get something out of RAM nodes. 
Everything else
> is not affected - persistent queue still persist, and transient 
operations are
> transient on disc nodes too.
> --
> Francesco * Often in error, never in doubt

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20120706/a7b153a1/attachment.htm>

More information about the rabbitmq-discuss mailing list