MQSeries.net :: View topic - Is it good practice to differ clustered remote q defs?

mqjeff · Posted: Tue Jun 04, 2013 6:05 am Post subject:

I think the point that Peter was getting at was that given that you have two sender/receiver channels outbound and those are not objects that the cluster is aware of ... then if one of those channels goes down, there's no mechanism of signalling the cluster to mark the relevant qremote as not available for cluster workload balancing.

One could potentially do things with channel events and scripting to alter the qremote...

It sounds like, instead, you're building a cluster that that only covers the two "internal" and two "external" end points, and then overlapping the two internal endpoints with an internal cluster. Then messages are workload balanced to the two internal endpoints by the internal cluster, and delivered across the boundary using the external cluster.

It's a topology that's relatively straightforward to draw out - a big circle with an intersecting smaller circle. But it requires a certain amount of rigorous planning out of each and every object. The power and pain of clusters is that they always want to fully connect with everybody. So if you do the wrong things in your setup, you've now got internal queue managers trying to talk directly to external queue managers.

Best bet is to view them as entirely separate clusters, and plan them as entirely separate clusters, and then view the overlap as a standard mulit-hop between three queue managers.