On 4/25/07, DavidR <[EMAIL PROTECTED]> wrote:
Hi, we are trying to use ActiveMQ clustering using store and forward.
If you want clustering (high availability and failover of brokers) you really should use Master/Slave instead of Networks... http://activemq.apache.org/masterslave.html
we see that when we use temporary queues and then try to delete the queue we get an error message: "Destination still has an active subscription". Browsing the queues in JMX, we see that the temporary queue message was only picked up on one of the machines and not the other and that seems to be the reason the delete is failing. Is this a bug?
It could be.
Additionally, is it necessary to delete the queue or can we rely on it getting cleaned up?
It should get cleaned up when the connection goes away
As for the configuration, we saw what James wrote that he did not recommend store and forward if in the same LAN but we do not see too many good possibilities: 1. JDBC master slave -- uses a lot of CPU on the database 2. regular Master Slave, as documented, is quite difficult to restore if
3. shared file system... Well if you don't want to replicate messages to another store, you don't even need networks-of-brokers - just running separate brokers is fine. The only thing networks of brokers buy you is the ability to move messages from one broker to another which you may or may not want. -- James ------- http://macstrac.blogspot.com/
