activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "James Strachan" <>
Subject Re: High Availability and Load Balancing
Date Wed, 21 Nov 2007 08:45:49 GMT
The reason for master/slave is to avoid catastrophic hardware failure
loosing messages. i.e. if you loose an active broker's hardware
machine (its disk or its data centre entirely) you've lost the
messages - with Master/Slave the slave is kept in sync so has all of
the messages the master had to avoid message loss in catastrophic

If you want all brokers to be active, just don't use master/slave.

On 20/11/2007, ktecho <> wrote:
> James.Strachan wrote:
> >
> > On 20/11/2007, ktecho <> wrote:
> >>
> >> In a store and forward network of master/slave clusters, all the brokers
> >> receive messages?
> >
> > All brokers in a master/slave cluster receive the message; only the
> > master dispatches it to consumers.
> >
> > Then messages are stored and forwarded to other master/slave clusters
> > as they pass along the network-of-brokers (if there are consumers
> > available on other master/slave clusters).
> >
> >> That would be nice, as that way we can have all the
> >> parsers from all the machines working and busy. That's why I was looking
> >> for
> >> Active/Active mode.
> >
> > I still think you're missing something. The broker just dispatches
> > messages to consumers. Consumer are the things which would parse the
> > message and do something with it. To a broker; a message is a just a
> > blob of data to be routed around a network - the Consumer is the thing
> > which actually does something with it.
> >
> Yeah, I understand that consumers are the entitys that process messages.
> Sure I could have only one broker in one machine and two consumers in two
> different machines that make the parsing, but I think it would be even
> better if I could have two active brokers with it's active producer (one set
> of broker/parser on each machine), so I have the load balanced between both
> machines. That way, if one of them fails, the other one can still process
> it's messages and the ones from the machin that contained the broker that
> failed.
> If I understood you correctly, using Master/Slave "store and forward", only
> one of the two brokers will dispatch messages (perhaps to both machines), so
> I loose the capacity of the other broker, that is in passive state, waiting
> for the Master broker to die. It's just that it would be nice if both
> brokers could work together as Masters, and be so friendly that if one them
> fails, the other one can finish dispatching its messages.
> --
> View this message in context:
> Sent from the ActiveMQ - User mailing list archive at


Open Source Integration

View raw message