Active MQ embedded broker; practical and reliable in production?

Question

Currently evaluating ActiveMQ for JMS transport between multiple sites.

Active MQ's embedded broker is attractive because most of our systems are Java deployed on Tomcat/WebLogic and, so the embedded broker sounds tempting, but we're not sure if this approach is practical in production. Our concerns...

• Is embedded broker stable and reliable?
• Is embedded broker functionally limited compared to standalone, remote broker?
• Does embedded broker work with broker network?

Not sure if we're missing anything else, just looking for some advice/pointers on this approach

Answer 1

We're using an embedded ActiveMQ 5.x in each RSB node (http://www.openanalytics.eu/r-service-bus) for dealing with work requests and results. This infrastructure is used in production in different context (financial, medical) and has proven reliable.

Even scenarios where the JVM hard crashes doesn't end up with a corrupted ActiveMQ local store.

I am not aware of reduced functionality in embedded mode (even JMX is available). An ActiveMQ specialist may chime in here and shed some light.

You could achieve store and forward by relying on Mule to perform the interconnections between your embedded brokers, using reliability patterns like consuming messages in transactions to drive redelivery attempts. You could also use ActiveMQ's network of broker: my personal experience with it was that it was not very stable in 4.x but I'm certain that it's now great in 5.x.