activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Hiram Chirino" <hi...@hiramchirino.com>
Subject Re: Getting started with ActiveMQ
Date Wed, 26 Jul 2006 22:23:17 GMT
On 7/26/06, jcarreira <jcarreira@gmail.com> wrote:
>
>
>
> Hiram Chirino wrote:
> >
> >
> > We have discovery agents and also a discovery transport.  The discovery
> > transport uses a discovery agent to find brokers and to connect to 1 of
> > them.  The peer protocol implicitly starts up an embedded broker that is
> > configured to use a discovery agent so that the broker interconnects
> with
> > the other peer brokers.   So, yes peer brokers uses discovery to find
> > peers.  Our discovery agent (multicast based by default) is only used to
> > discover other brokers.  The communication between nodes always uses
> tcp.
> >
> >
>
> Is there any more documentation on peer setup than
> http://incubator.apache.org/activemq/peer-transport-reference.html ?


yeah that's all we have so far :(


Hiram Chirino wrote:
> >
> >
> > Here is a bean that creates a broker that is configured using a
> > activemq.xmlin the classpath.
> >
> >     <bean id="brokerContainer" class="
> > org.apache.activemq.xbean.BrokerFactoryBean">
> >       <property name="config" value="classpath:activemq.xml"/>
> >     </bean>
> >
> > And here is a bean that creates an ActiveMQ connection factory to the
> > broker.
> >
> >     <bean id="factory"
> > class="org.apache.activemq.ActiveMQConnectionFactory
> > ">
> >         <property name="brokerURL" value="vm://localhost" />
> >     </bean>
> >
> > Note that broker url must match one of the connectors configured for the
> > broker.  By default, all broker can be accessed using the
> > "vm://${brokerName}" URL, and the default brokerName is always
> "localhost"
> >
>
> the "vm://${brokerName}" syntax works no matter how the broker is
> configured? It just automatically maps to a broker by that name configured
> in this VM?


Right.

Hiram Chirino wrote:
> >
> >
> > All the brokers can share the SAN.  If you set them up to use the same
> > journal files on the SAN, then only 1 of the broker instances will be
> > fully
> > started.  The other instances pointing at the same journal files will
> wait
> > for the master to fail before taking over the journal (and fully
> starting
> > up).
> >
> >
>
> Ok... Is that the recommended way of doing this? Seems like it will limit
> throughput to whatever the master machine can push to disk and scaling the
> cluster won't scale up. Not that it really matters for my usecase, since
> it
> will be pretty low traffic...


Most users find SANs to be a little too expensive so they go with
http://activemq.com/site/masterslave.html setup so that the data replication
is done by the brokers and therefore a SAN is not needed.  This also avoid
the SAN becoming the bottleneck.  It has it's pros and cons.

There are so many different use cases and different deployment scenarios
that it's hard to pick a 'recommended' way to configure your brokers.  We
aim to stay flexible and offer lots of options so that you and pick what's
most optimal for your scenario.


Thanks very much for the help Hiram!!
> --
> View this message in context:
> http://www.nabble.com/Getting-started-with-ActiveMQ-tf2004874.html#a5509287
> Sent from the ActiveMQ - User forum at Nabble.com.
>
>


-- 
Regards,
Hiram

Blog: http://hiramchirino.com

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message