Master/Slave Round Robin of Brokers

[MrEcho <mr...@gmail.com>]

Im trying to create a situation where there is no single 'master' in the
cluster of brokers. Single point of failure.

I need to use master/slave method because of the HA that I need. But if the
'master' goes down, I can NOT just stop the slave and copy over /data. Ive
got live udp packets of stuff coming in that I cant buffer in my application
till the master comes back online. From the looks of it the single /data
file only 1 borker can have access to it, and thats just to risky for me.

So my idea is to have 3 brokers, each being a Slave to the next, but also
being a master. Dual Ring round robin kind of thing. One goes down, but
because its a slave, the master should send all the stats/connection data to
the one that was down. And its slaves reconnect to it. The clients would
have failover: for all 3 brokers.

Is this even possible? What about race conditions? Has anyone done this?

-- 
View this message in context: http://activemq.2283324.n4.nabble.com/Master-Slave-Round-Robin-of-Brokers-tp2718127p2718127.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.