Clustering - Load balance and failover together - any 7x24 solution?

[Gordon Luk <lu...@ha.org.hk>]

Dear All,

I am new to ActiveMQ. I found ActiveMQ 5 support failover custering (Master
/ Slave). Master data replicate to slave and had limitation*. Other options
are shared file system and database, two options also could appear single
point of failure at backend side. Sure that, we could setup failover support
on the backend side. How can i setup 7x24 service? thanks.
-- 
View this message in context: http://www.nabble.com/Clustering---Load-balance-and-failover-together---any-7x24-solution--tp16671477s2354p16671477.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.

Re: Clustering - Load balance and failover together - any 7x24 solution?

[Abdul Alhazred <th...@tradedesksoftware.com>]

Gordon Luk wrote:
> 
> Dear All,
> 
> I am new to ActiveMQ. I found ActiveMQ 5 support failover custering
> (Master / Slave). Master data replicate to slave and had limitation*.
> Other options are shared file system and database, two options also could
> appear single point of failure at backend side. Sure that, we could setup
> failover support on the backend side. How can i setup 7x24 service?
> thanks.
> 

There should be a few possible ways to accomplish this. The 1st observation
I would make is that with modern HA infrastructure you should be able to
build either an HA RDBMS or file system that is reliable to the 'unless the
building burns down' level. So using shared FS or shared RDBMS with failover
should give you effectively the highest reliability you can achieve. Granted
these are not cheap or simple things to set up, but if you want 24x7 you ARE
going to pay for it...

The other option would be to create a straight cluster of brokers. If you
have 2 brokers and peer them in a cluster, and use the failover transport,
then any broker crash should result in all clients on that broker seamlessly
failing back to another member of the cluster and messages will be
delivered. Depending on the exact usage it might need some configuration to
get it working for all possible cases. However it seems to be the most
straightforward solution, and has the virtue of not requiring a lot of fancy
HA setup for your message stores. Plus you can bring a failed broker back
into the cluster after you recover it, which you can't do with some of the
other solutions. 
-- 
View this message in context: http://www.nabble.com/Clustering---Load-balance-and-failover-together---any-7x24-solution--tp16671477s2354p16925823.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.