activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Hervé BARRAULT <herve.barra...@gmail.com>
Subject Re: ActiveMQ and memory cache
Date Tue, 24 Jan 2012 10:02:59 GMT
Hi,

Thanks for the answer (i agree with it).

Today, if i don't say something wrong, there is no way to share a
persistent adapter between multiple brokers. I only see Master/Slave
solution.

I have seen something with a broker cluster :
http://activemq.apache.org/clustering.html#Clustering-Brokerclusters.
But it seems to need a Store and forward networks of brokers which is
not an HA solution as written in the documentation.

Is there another way to have a multi master activemq broker ?

Regards
Hervé

On 1/23/12, Gary Tully <gary.tully@gmail.com> wrote:
> That would be great to spike into a proof of concept. The memory
> persistence adapter would be an ideal start point as it maintains some
> simple in memory lists that would be easy to replicate.
> Reusing something as solid as memcached makes sense if it is not too
> difficult to setup and maintain.
> For sure there are some usecases that would benefit from a simple
> shared memory persistence impl. However dealing with message order and
> concurrent consumers on the same destination will require some
> thought.
>
> On 23 January 2012 08:51, Hervé BARRAULT <herve.barrault@gmail.com> wrote:
>> Hi,
>> I have a question about an idea of deployment.
>>
>> Is it possible to use something like memcached to create only one
>> broker over multiple hosts/JVMs ?
>>
>> The main idea is to use an "embeded broker" in each JVM but seen as
>> only one broker. By this way, i can use VM protocol (no serialization
>> i  guess) instead of STOMP or TCP.
>>
>> In normal case the message will be produced and consumed in the same
>> JVM but only in case of a JVM/host crash, the message will be consume
>> in another one.
>>
>> I think also by this way to avoid persisting messages in files or
>> database but in a "shared distributed memory" ( it is not as safe as
>> persistence but i could loose messages only if all JVM crashes).
>>
>> Thanks for answers
>>
>> Regards
>> Hervé
>
>
>
> --
> http://fusesource.com
> http://blog.garytully.com
>

Mime
View raw message