I have added the logs with the traces to AMQ-2135: https://issues.apache.org/activemq/secure/attachment/17713/testamq1.log https://issues.apache.org/activemq/secure/attachment/17714/testamq2.log https://issues.apache.org/activemq/secure/attachment/17715/testamq3.log
Attached 3 log files from running the example provided in the initial issue (with log4j.logger.org.apache.activemq=TRACE). The result of the test is that the first consumer (connected to 61616) consumes 5 messages, the second consumer (connected to 61617) consumes 2 messages, and 3 messages are orphaned on the 3rd broker (61618). Given that this very simple use case of a network of brokers is failing, I have to ask: Is anyone actually using this functionality with success? Could you please share your configurations? Thanks, Garth On Tue, Mar 3, 2009 at 9:01 AM, Garth Patil <[email protected]> wrote: > Thanks Gary. > Using the configuration in AMQ-2135 and turning on TRACE logging > level, I'm able to see messages going to broker 3 and being orphaned > there. I have also tried a permutation of that configuration that sets > conduitSubscriptions="false", and I get the same result. > Also, per this thread, it appear that this is happening in 5.2 and not 5.1: > http://www.nabble.com/ActiveMQ-5.2-Network-of-Brokers-bug-td22280154.html > Best, > Garth > > On Tue, Mar 3, 2009 at 8:50 AM, Gary Tully <[email protected]> wrote: >> Hi Garth, >> No. A consumer is per destination. It is active on creation till it closes >> or its session closes or its connection closes. The relevance of the >> connection relates to the inactivity timeout which may kick in during a long >> ack window. >> >> The network support is based on advisory messages so a networked broker is >> aware of per destination consumer creation. >> >> I added a comment/question to AMQ-2135. >> >> Gary. >> >> 2009/3/3 Garth Patil <[email protected]> >> >>> So an active consumer is per connection rather than per destination >>> (queue/topic)? In the case of a Network of Brokers (with >>> dynamicOnly=true), this definition seems critical, because messages >>> should only be forwarded to remote brokers if there is an active >>> consumer on a specific destination. >>> From http://activemq.apache.org/networks-of-brokers.html >>> dynamicOnly = if true, only forward messages if a consumer is active >>> on the connected broker >>> >>> Is this what is going on in this bug: >>> https://issues.apache.org/activemq/browse/AMQ-2135 >>> where messages are being dispatched to brokers with no consumers >>> because the transport connector of the remote broker itself is being >>> considered an appropriate "active consumer"? >>> >>> Thanks, >>> Garth >>> >>> On Tue, Mar 3, 2009 at 5:31 AM, Gary Tully <[email protected]> wrote: >>> > 2009/2/25 Eric Chu <[email protected]> >>> > >>> >> Hi everyone, >>> >> >>> >> 2 part question: >>> >> >>> >> 1) What is the exact definition of an "Active Consumer"? I'm wondering >>> what >>> >> conditions would cause a broker to consider that a consumer went from >>> >> active >>> >> to inactive. >>> > >>> > >>> > In ActiveMQ, in the absence of a close() call, 'active' is tied to the >>> > connection. >>> > >>> >> >>> >> 2) Specifically, if a consumer receives a set of messages and does not >>> send >>> >> back an ACK for a period of time (let's say 5 min), is it still >>> considered >>> >> active? >>> >> >>> > >>> > Yes, but you may encounter an inactivityTimeout on the connection if >>> using >>> > tcp: and not failover. If a connection is inactive for more than 30 >>> seconds >>> > (by default) it is considered dead and will be closed. See the reference >>> for >>> > details on how to configure an >>> > inactivityTimeout< >>> http://activemq.apache.org/configuring-wire-formats.html> >>> > >>> > >>> > >>> >> Thanks in advanced! >>> >> Eric >>> >> >>> > >>> > >>> > >>> > -- >>> > http://blog.garytully.com >>> > >>> > Open Source SOA >>> > http://FUSESource.com >>> > >>> >> >> >> >> -- >> http://blog.garytully.com >> >> Open Source SOA >> http://FUSESource.com >> >
