activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From nlif <>
Subject Problem using network-of-brokers and auto-discovery
Date Sun, 22 Jul 2007 15:16:05 GMT

Hi all,

We're having problems running a cluster of brokers (AMQ 4.1.1) with
auto-discovery. We're using auto-discovery via multicast both for the JMS
clients and for the network-of-brokers. Our architecture is as follows:

There are 4 clients, and 2 brokers. The clients are configured like this:


The brokers are configured like this:

       <transportConnector name="TPS1" uri="tcp://"

     <networkConnector name="TPS" uri="multicast://"/>

Now, here's what happens: when we start all processes, all work fine. Next,
we kill one broker. Everything is still fine. Presumably, the clients that
were using the broker that was terminated, switch to the live one. The
problem occurs once we restart the broker that was terminated. From this
moment - the system starts behaving strangely: Some requests work, some
hang... we cannot find a pattern to the problem.

What we do know, is that once we replace the discovery with a hard-coded URL
list, things work fine (when killing and restarting one of the brokers).
Like this:


Has anyone seen such behavior? Any ideas? Thanks in advance.

Addition: after some more testing, we found another problem. I'm adding it
to this question, because the configuration is the same...

When using the second configuration described above (i.e. the one with the
hard-coded URLs), we found that if we disconnect the network cable of the
machine on which one of the brokers runs, the system cannot recover. We
expected that it will - after all, this is a valid failover scenario. We
don't know whether it's an AMQ issue at all. Maybe it has to do with the TCP
settings of the machine (centOS 4.4).

Thanks in advance,

View this message in context:
Sent from the ActiveMQ - User mailing list archive at

View raw message