have a peek at https://issues.apache.org/jira/browse/AMQ-2807, it is an implementation of your idea using a local file or any uri provider
On 15 November 2011 09:15, Radek Kraus <[email protected]> wrote: > Hello, > I running something like cluster application. I used "embedded" ActiveMQ > broker, which can run only as one instance (singleton) in whole cluster app. > If the "cluster node" with active ActiveMQ broker died, then another > ActiveMQ broker is started => so I have only one "active" instance of > ActiveMQ broker at one time (JDBC persistence). > > On the "client side" I used failover transport at first (I must specify all > cluster node addresses, where ActiveMQ broker can run - failover uri). I > would like to modify "static" failover configuration (all nodes must be > present in uri) to a dynamic way. > > The first simple idea is to write the uri of active ActiveMQ broker to JNDI > (or somewhere). The client then can ask JNDI for active broker uri. > If the connection failed, client ask JNDI for current active broker uri. For > me, it sounds as "modified" failover transport. > > I want to ask if the idea to write own custom transport (based on failover > transport) is correct. Or if there is another way how to reach this goal. > > I cannot used none of "clustering" features, which is available in ActiveMQ > (Networks of brokers ...) at the moment, let say that it is a requirement to > use singleton broker in own cluster architecture. > > Thanks in advance. > > -- > View this message in context: > http://activemq.2283324.n4.nabble.com/Custom-Transport-similar-to-Failover-transport-tp4042314p4042314.html > Sent from the ActiveMQ - User mailing list archive at Nabble.com. > -- http://fusesource.com http://blog.garytully.com
