I'd like to know what u people think about the conception I created.
I want to build a cluster of 'nodes' - a sort of platform we built.
Each node starts its own embedded broker on startup. The embedded brokers
find each other using discovery
broker = new BrokerService();
broker.setBrokerName(values.getName());
broker.setUseJmx(values.isJmx());
broker.setPersistent(values.isPersistent());
try {
connector = broker.addConnector(values.getUrl());
connector.setDiscoveryUri(new
URI("multicast://"+multicastAddress));
broker.addNetworkConnector(new
URI("multicast://"+multicastAddress));
connector.getDiscoveryAgent().setBrokerName(values.getName());
} catch (BindException be) {
// address is in use already
}
and
connector.start();
broker.start();
Each node has a queue through which it should communicate to other nodes.
To find each other each node sends some sort of 'I am alive'-message
(containing a node-queuename) into a globally specified topic. It listens on
that topic too.
Every node then sees other nodes and can get the queuename of other nodes.
Is it generally advisable to have a network of embedded brokers?
Is there anything which must be taken care of ? E.g. brokernames?, special
setup?
In practise this works on one computer, but causing troubles when using
different computers.
(in Jmx-console: broker1 can see the broker2, but doesnt get messages;
broker2 cannot see the broker1 but does get messages from broker1)
Any hints would help me.
--
View this message in context:
http://www.nabble.com/CommunicationClustering-based-on-ActiveMQ-tf3645535s2354.html#a10181323
Sent from the ActiveMQ - User mailing list archive at Nabble.com.
