[ivelin.ivanov]

I have recently commited a few features to the clustering framework (Singleton and Scheduler) and would like to understand whether it is worth it implementing JMS clustering in Branch_3_2 ?

Is the 4.0 code so different that such a contribution would be thrown away in a couple of months?

If it is worth persuing this, how long do you think it will take and what is a good starting point ?

There is also a related thread on the clustering forum.
http://jboss.org/thread.jsp?forum=155&thread=34188

[adrian.brock]

Hi ivelin,

I saw your commit about the HASingleton.

I also commited a jdbc state manager in 3.2
It stores the state to the db rather than the
file system. This means we can also failover
durabe subscriptions.

You can use the database server login module
for jbossmq security in login-config.xml

Coupled with your HASingleton, this would
provide automated cold failover of jbossmq.

If you want to write the singleton service
that manages the failover, let me know.

The idea being jbossmq only starts on one
server, if that server goes down, another
server takes over.

The big disadvantage is that the db is a single
point of failover. Of course, the db can
be replicated.

For jboss4 we want to be able to start
messaging on all servers. The singleton service
still makes sense if you want to avoid the
overhead of replication. e.g. You might only start
a queue/MDB on one server with cold
failover.

Regards,
Adrian

[ivelin.ivanov]

Sorry to read your response so late.
I will email you offline about the implementation.