I’m reading this forum and in few post somebody mentioned that he has active/active broker cluster deployed. I have some experience with wM but still I thought that this is impossible to achieve. I know that wM treats territories/gateways as a way to scale broker architecture but maybe there is also a different way to achieve that ?
In my scenario there is whole IS farm connected to single broker. Also these IS’s running as a cluster (same serverId) in order to run PRT processes. Any single step of any process can be executed on any IS. Transition between steps are transfered through broker (TransitionDoc).
In that scanario deploying territories/gateways will limit a possibility of distributed step execution or am I wrong with it ?
There is no such active/active cluster solution at all, however in version 8 a new feature for JMS client is available. The client keeps information on several Broker servers and apply different load balance policies.
Therefore the JMS clustering is implemented on “client” side, and is only available for JMS.
The parts of wM which work with native Broker will not have this.
The different policies you can configure for the clients are explained on page 392.
You can have several Brokers in active/passive ( The so called HA ), and have JMS clients with clustering enabled.
However is true ( only for JMS without XA and not native Broker ), if you only make use of JMS and they have policy Multisend Best Effort, you can achieve somehow a active/active solution. Keep in mind this policy is not free… it has a cost of publish to all brokers so probably will slow down throughput.
I recall the active/active Broker configuration thread from a few years back. Alas there were no details about how it was done and what, if any, limitations there were. Brokers in active/active is not supported by wM/SAG, so probably best to avoid that and explore the approach DevNull outlined above.