Others have seen similar behavior; for example, see http://activemq.2283324.n4.nabble.com/ActiveMQ-master-slave-topology-issue-BUG-td4695677.html#a4695731 and https://issues.jboss.org/browse/ENTMQ-975.
I don't know whether a fix of any kind has been identified/implemented. On Thu, May 28, 2015 at 8:10 PM, clivewang <[email protected]> wrote: > I created a network of brokers with two broker: broker1 and broker2 on two > separate vm. > then created two slave brokers to backup brokers, each backup the master on > the other vm. > when unplug the network cable of one vm of broker1, the slave broker on vm2 > will take over the KahaDB and become master after about 90 sec. > Then plug backup the network cable of vm of broker1, there will be two > master brokers using the same KahaDB store. and each broker seems to be > independent broker. each broker can receive/send msg independently( they > should be one master and one slave ). but if you stop/start any of the > original master or escalated master, the msg on that master is lost and > become slave. > > > > > > -- > View this message in context: > http://activemq.2283324.n4.nabble.com/Fail-back-not-correct-after-network-interruption-tp4697031.html > Sent from the ActiveMQ - User mailing list archive at Nabble.com. >
