> Hi,
> My question is, if a broker in cluster setup fails,
> do the other brokers know about the messages sitting
> on this failed broker?
> If so can one of the other brokers send them out to
> relevant consumers?
> I would very much appreciate your reply.
>
In the current release, a cluster allows scaling of connections
and availability of service but does not support high availablity
of message data. This means that connections can fail over
to a new broker and continue to send or receive messages,
but the messages stored on a down broker (aka the broker
that the producer was connected to) will remain unavailable
until either:
* the broker restarts
* a new broker is started which points to the same store
If you need the message data to be highly available, we
do support that functionality through SunCluster on solaris.
(sun cluster handles storing the data on a highly available
file store and automatically starting a new broker to take over
for a down broker)
We are also currently working on adding HA functionality to
MQ so that we can provide this functionality in a future
release
Thanks for your reply.
I am thinking to try and resolve this by plugging in an external database such as oracle (as a shared persistance) and configuring all the brokers to connect and use this database as their message storage area. In this way all brokers will be looking at one data source for sending and storing messages.
So if one broker fails the other broker does not need to know what messages are sitting on this failed broker and basically look at the shared database to check what messages to send.
In your view is this possible?
Will the brokers work with a shared database between them?
Appreciate your reply on this.
Thanks