There are a few more observations I've made while trying to debug this. We've noticed that sometimes the messages received from the consumer queues have JMSDestination set to the VirtualTopic name, and at other times they have the consumer queue name. As shown earlier, I do have the excludedDestinations clause necessary to exclude forwarding to the queues. Is this still to be expected?
Also, looking in JMX at the VirtualTopic's topic MBeans, I've noticed that the topics themselves have consumers in addition to the queue. Are these consumers created internally by the broker to enable forwarding, or could this indicate a problem? Further, there are far many more consumers for the topics. For example, when testing with 2 nodes: Although I have only one consumer queue attached to the topic, and one consumer to this topic on each node, I notice that the VirtualTopic itself shows 6 consumers. Where are these consumers coming from? Another observation when looking at the MBeans: the QueueSize of the topics appears to increase steadily over time. When I try to browse the contents of the topic, I often find that there are some residual messages (usually far fewer messages than indicated by the QueueSize statistic, but > 0 nonetheless). Again, this doesn't gel with my understanding. I thought that the topics themselves are "virtual" which should imply that they don't have a persistent cache of messages anywhere? Are the virtual topics themselves also persisted? If so, when do messages in the topic get evacuated? If someone familiar with this area could answer these questions, it could really help us make some important decisions about how, and whether, to use virtual topics to implement our requirements. Has someone used them effectively in a network-of-brokers scenario? If so, can you share the specifics of your configuration? -- View this message in context: http://old.nabble.com/Problems-with-Virtual-Topics-tp28100311p28107798.html Sent from the ActiveMQ - User mailing list archive at Nabble.com.
