Messaging Cluster scenarios

[kapilanand]

I have two questions regarding the messaging cluster. I could not find the answer in the documentation.

This is what I need to achieve:
Jboss messaging cluster to provide seamless fail-over for the publishing and subscribing clients.
Persistence and durable subscription are not needed. In other words, an unreliable but highly available replicated topic.

1. HA and load-balance without persistence
Messaging doc says that a shared DB is required for the cluster to work. Before that it states that the post-office will not work without a DS. I am aware that embedded DB may be configured to avoid persistence until a threshold of memory usage is hit.

My question related to this:
- Would I need the shared DB even if I use these clustered post-offices to host non-persistent (non-durable) subscriptions, that is, is it used to manage the cluster information as well?
- Is it OK to use HSQLDB or Derby if the above is not true?

2. Recovery from a cluster restart
Since Jboss messaging relies on a clustered ConnectionFactory registered in a clustered JNDI to provide the failover mechanism, would the client need to re-lookup JNDI and reopen connections and sessions to get back in business? This seems unnecessary since all participating post-offices seem to be bound to a pre-defined host and port, but then I am not sure how it is different from a clustered Stateless session bean.

thanks
kapil

[kapilanand]

Could somebody please point me to the right documentation that answers my questions?

On JBoss Wiki I ran into few articles related to JBossMessagingCore which gives good description but not the actual configuration step. It's perhaps because it is not meant to be used directly.

What I am looking for is equivalent of ActiveMQ peer-to-peer transport without persistence.

[kapilanand]

I am sure this message has been read by the experts of the matter. Please let me know if posting this via JBoss Network subscription will help me in getting the answers.
Of course, last resort for me would be to do some experiments which I guess are inevitable even if I get some hints and tips from the forum.

[ataylor]

My question related to this:
- Would I need the shared DB even if I use these clustered post-offices to host non-persistent (non-durable) subscriptions, that is, is it used to manage the cluster information as well?
- Is it OK to use HSQLDB or Derby if the above is not true?

You need to use a shared database for clustering, HSQLDB is not usable for a clustered configuration.

2. Recovery from a cluster restart
Since Jboss messaging relies on a clustered ConnectionFactory registered in a clustered JNDI to provide the failover mechanism, would the client need to re-lookup JNDI and reopen connections and sessions to get back in business?

No, this will all be handled transparently by the client.

Please let me know if posting this via JBoss Network subscription will help me in getting the answers.

yes, we're just having a busy week. :)