There will be no JMS clustering support for 3.0. Sorry.
when such support should be added?
Which kind of clustering support do you expect for JMS? Which specific features?
Yes, it would be cool to get some user feedback on features they want for JMS clustering.
Hi Bill, it's mee, pra, you know ;-)
I would verry much like to have JMS clustering and have been longing for it for a verry long time.
I think the most important aspect would be HA. I.e you create a JMS connection and publish or subscribe to a destination. This destination should be failsafe when clustered in regard that the client will be redirected to a new node if the current one fails.
At the lowest level this ought not to be to complicated by making a JavaGroup clustered queue. The problem, i would guess, has to do with the persistence and TX constract for JMS.
By the way, what are the org.jboss.mq.il.ha packet for in the cluster prorject?
In an attempt to bridge all the JMS clustering/federation players I've created a topic in the JMS forum. Seems that most work will probably be done by these folks so this type of "clustering" might best be discussed there.
> By the way, what are the org.jboss.mq.il.ha packet
> for in the cluster prorject?
It's a HA version of an IL for JBossMQ. So basically, if you get HA IL connection and the node your connected to goes down, you'll transparently be reconnected another node.
So I did the easy part. Now who is going to do that hard part of making a JMS destination reside at multiple nodes???
What about persistence?? Should it be centralized or should it be distributed?
Well, the persistence should probably be distributed. But that is not in itself necessary the hardest part. Is it not the infrastruktur of publishing itself:
1. How do we govern that only one consumer is allowed to get a message from a distributed queue, is it possible to synchronize it globaly?
2. How do we remove sent messages from topic? It does not seem a good idea that each node remove the message from the topic when is's subscriber have received it, but who does.
3. How do we do about recovery? If one node goes down, with messages still in its queue. When it commes up again and recovers from persitant storage, how does it know wich messages was actually allready delivered by other live nodes.
4. How do we get the transactions to span all nodes, can we use the distributed TX manager?
There we have some really tough issues from an architectural point.
I have a problem for many JMS Server.
When I have many LANs, I hope there is a JMS Server per LAN.
When message communication is in LAN, we use local JMS Server.
But When I want send a message to the bean in the other LAN, what can I do?
How can I process many JMS server in WAN?