Hi Gary, Thanks for the reply. I've tried to attach the activemq.xml - for some reason it's been filtered out. I do have the updateClusterClientsOnRemove="true" attribute on the transport connector. Sorry I did not list it in the original email.
Here is the snippet. Broker 1: <networkConnector uri="static://(tcp://host2name:62616)" conduitSubscriptions="false"/> <transportConnector name="openwire" uri="tcp://0.0.0.0:61616" updateClusterClients="true" rebalanceClusterClients="true" updateClusterClientsOnRemove="true"/> Broker 2: <networkConnector uri="static://(tcp://host1name:61616)" conduitSubscriptions="false"/> <transportConnector name="openwire" uri="tcp://0.0.0.0:62616" updateClusterClients="true" rebalanceClusterClients="true" updateClusterClientsOnRemove="true"/> Will take a look with the code. Not sure which module triggers the re-balance. Thanks, again ________________________________ From: Gary Tully <gary.tu...@gmail.com> To: users@activemq.apache.org Sent: Mon, February 28, 2011 7:28:49 AM Subject: Re: Connection Re-balance Question Just had a peek at the code[1], you need to add updateClusterClientsOnRemove="true" to your transport connector. It looks like the cluster info is not aware of the broker failure (node removal) and as a result does not update then it joins again. org.apache.activemq.broker.region.RegionBroker#removeBrokerInClusterUpdate [1] http://svn.apache.org/viewvc/activemq/trunk/activemq-core/src/main/java/org/apache/activemq/broker/region/RegionBroker.java?view=markup On 25 February 2011 22:14, Joe Smith <joesmithc...@yahoo.com> wrote: > Hi, > > I'm using the 5.4.x transport's updateClusterClients="true" > rebalanceClusterClients="true" options. > > > It works fine as a broker is added to the cluster - the existing connections >are > rebalanced. > > > When a broker that's already in the cluster when down, the connections failed > over successfully to the other broker (clients are using failover://). > However, > when I bring up the failed broker, the connections on the other broker are not > re-balanced. > > My questions are: > > 1. Does connection re-balancing work in this scenario - where a known broker in > the cluster comes back on line after a crash? > > 2. What criteria need to be met before re-balance would occur other than when a > NEW (vs. existing) broker is added to the cluster? E.g. we have only 24 client > connections, is that too low? > > We tried both 5.4.1 and 5.4.2. Linux. JVM is 1.6. > > Thanks for the help. > > > -- http://blog.garytully.com http://fusesource.com