Couple of things: - It looks like your cluster is quite a bit larger than 3 nodes. By my count, there are 12 nodes - 6 lives and 6 backups. Can you confirm? - The functionality described in the documentation you cited [1] is meant to be used when you don't have enough live/backup pairs to establish a quorum in order to mitigate split-brain. Since you have enough nodes to establish a quorum you don't need to configure this. You should allow the nodes to vote and make a decision that way.
If you ever find that a live and it's backup are active simultaneously then the only way to resolve the situation is to stop at least one of them. Depending on the circumstances it may be best to stop both of them in order to determine which has the "best" data. Recovering from a split-brain is not something brokers can do on their own. It requires administrative intervention. Justin [1] https://activemq.apache.org/artemis/docs/2.4.0/network-isolation.html On Thu, Mar 28, 2019 at 10:39 AM adjebbi <djebbi.ach...@scheidt-bachmann.de> wrote: > We have a cluster with 3 nodes : > > Amsterdam (2 live / 1 backup) > Frankfurt (1 live / 3 backup) > Paris (3 live / 2 backup) > After isolating the network of Amsterdam, Paris backup server takes over > the > cluster. all works fine. But when Amsterdam joins the cluster we observe > that we have 2 live servers (Paris and Amsterdam) and the backup still in > Paris. We have tried the configuration like describe in the following link > https://activemq.apache.org/artemis/docs/2.4.0/network-isolation.html > without any success the resolve this issue. > > The only solution that we have until now is restarting Amsterdam live > server > and Paris backup server. > > > > -- > Sent from: > http://activemq.2283324.n4.nabble.com/ActiveMQ-User-f2341805.html >
