i was wondering about messaging in a clustered environment, i come from jboss 4 and what i was used to do, is to use a DBMS to store the queue and have N active consumers... the master node switches automatically on failover and everything seems to be ok...
HornetQ stores queues in the file system... so, every node has its own queue, this addresses load balancing, failover is addressed with N passive nodes that shares the file system with the active ones.
So, when i set up an architecture with jb4, i can prepare 2 virtual/phisical nodes and the cluster is up both for balancing of the listeners and for failover, with jb6/7 (hornetq based), how can i set up that architecture?
2 nodes are not enough, i must sacrify or balancing or failover, i need at least 4 nodes?
Application server up in every node, hornetq up only on 50% nodes?
filesystem is shared between running and backup servers, what happen if i share the file system between the active ones? In particular if i configure the client to balance the publisher?
sorry for my bad english
I have more or less the same doubts.
Is is possible to have HA in Active/Active mode by sharing storage? If so, do I need a special filesystem (one that supports concurrent access, like GFS).
To put in other words, while keeping HA, can I scale by adding one single node at a time?