On Mon, Mar 30, 2015 at 2:09 PM, wonderkind <kevin...@am.sony.com> wrote: > How big is your network? Do you see any significant degradation with a > network of replicated levelDB brokers?
I haven't started stress testing the configuration yet, the current plan is to have network two clusters of three. So six machines in all, only two active as the master at a time. > [ksu] How are you planning on discover the brokers? Via multi-cast > auto-discovery, or by specifying URLs statically? The later gets harder to > maintain as the the fabric’s membership grows. However, I am not sure if > there is any timing issue associated with the auto-discovery of masters, and > the auto-discvoery of brokers for the network. I don't see any reason why auto-discovery wouldn't work if you wanted to go that route. From the rest of the codes perspective there's one active broker per cluster, so I would think it would work just like networking stand-alone brokers. I'm not using auto-discovery myself, I'm planning on following naming conventions and having my cfengine system maintain the configuration, building a complete graph of the network. I've no idea how well, or poorly, that will work. You can see some of my configuration details by looking at a previous email I sent to the list: http://mail-archives.apache.org/mod_mbox/activemq-users/201503.mbox/%3CCAPd04b7106shHXUR+1VUddBmPUH+=mp414xoxnnpeuoukwf...@mail.gmail.com%3E > [ksu] Could you share the patch? Which version of ActiveMQ are you using? The patch is listed in the ticket. One of the maintainers did respond to the ticket, indicating he was going to look it over and incorporate it if it passed muster. Jim
