This is to follow up on http://www.nabble.com/question-about-the-store-forward-algorithm-to14223478s2354.html (but reading that isn't necessary).
Let's say I have a big graph of brokers connected to each other and producer A on broker A' sending a message to topic T. Further, let's say I have consumer B subscribed to T on broker B', which is a few brokers away from A' in the graph. Let's also say that NO other brokers in the graph (even those on the path(s) between A' and B') have any consumers subscribed to topic T. I know the message will get to B if I set networkTTL to be high enough. (I tried it, and of course it works.) My question is, in this process, where does the physical message actually get sent over the network? (Again, I realize that, regardless, it will get everywhere it NEEDS to go, but the question here is will it also go places it does NOT needs to go.) Does it go to EVERY broker in the graph (regardless of who's subscribed to what, if anything), or does it only go through the shortest path of brokers between A' and B'? If the latter, how is that accomplished? Does every broker in the graph KNOW which remote brokers around it are subscribed to which topics, so it determines where to forward the messages based on that? This will be useful for designing our (potential) distributed network of ActiveMQ brokers. It will let us estimate the amount of traffic that will be going through the system. P.S. In the above link, raj says that if EVERY broker in the graph has a consumer subscribed to T, then the message is physically sent everywhere in a BFS (breadth-first search) fashion. But the above question is different: what happens when only SOME brokers have consumers subscribed to T. -- View this message in context: http://www.nabble.com/another-question-about-networks-of-brokers-tp14407695s2354p14407695.html Sent from the ActiveMQ - User mailing list archive at Nabble.com.
