After some more digging it seem to be the failover transport that might be
the culprit. When posting a msg directly to one node
tcp://server_1:61616
the filter is applied correctly. But using the failover transport when
posting the msg
failover:(tcp://server_1:61616,tcp://server_2:61616)?initialReconnectDelay=100
results in the msg being forwarded from the VirtualTopic to the Queue, with
the selector *only* applied randomly.
Don't understand how the transport can have such an impact on the delivery
of msg to correct queues?
The NOB config is according to:
<networkConnectors>
<networkConnector name="a-2"
uri="masterslave:(tcp://server_2:61616,tcp://localhost:61626)"/>
</networkConnectors>
and correspondingly on the other node. So, in addition it's a
master/slave-setup. The setup seem to work fine, but during the tests the
slaves have been shut down. Creating a Queue (tied to the VirtualTopic) on
one node, automatically gets created on the other node.
VirtualTopic config:
<virtualTopic name="VirtualTopic.Customers" prefix="*."
selectorAware="true"/>
In addition the virtualSelectorCacheBrokerPlugin is used, as mentioned.
--
Sent from: http://activemq.2283324.n4.nabble.com/ActiveMQ-User-f2341805.html