Hey all, Is this a better place than the jboss forums to get help?
Anyway, I am most impressed with the feature set of JBoss clustering. Bill/Sacha did a great job. There are some issues which I have come across though and would like to point out, and possibly get help on if I am doing it wrong. So lets start off on the list: 1) Hot-farming works great, but only on already running nodes, and only if you drop in the .jar for the first time. When I re-deploy the .jar into one node, it redeploys on that server/node, but doesn't farm out to the other nodes. Thus, the only way currently to redeploy is delete (or move) the .jar out of a node's farm dir, then put in the new version. I don't see that this would be a hard fix. Some sort of file comparison in the url checking routine that already works when you drop a file in the farm dir (or deploy dir). The code is already there to redeploy, so I am not sure why it wont update the other nodes. But worse, is if you bring up a new "clean" node. I actually think i saw a message saying it was looking for node deployments, but I either get an error indicating unfinsihed deployment, or nothing at all occurs. Thus, if I bring on a new fresh node to "expand" my cluster, I have to undeploy/deploy the app for the new node to get it via farming. This should be part of the normal process.. that is a new node comes online, it automatically picks up any farming and/or deploy files. 2) What is farm-sevice.xml really for? We have seen that if we drop a file in the deploy dir, it gets farmed out. Yet, if we put farm-service.xml which seems to only indicate the ./farm dir, the farm dir works. I don't have farm-service.xml in my deploy folder and the farm directory still works. So I am a bit confused as to the real purpose of this xml config file, if it is still used, or does JBoss just automatically deploy anything in deploy and if a farm dir exists, that as well? Some things we have learned and might help others to watch out for: We have been testing our client app with three nodes at one time. We were seeing very odd round robin behavior and finally realized that our single button click was making several ejb calls. Also, we had output in only one part of the ejb side, and not everything, so we were mislead in not seeing all the possible output that was going on. What I advise is that if you are testing an app, watching the console output of your nodes, make sure you have verbose, debug and any other mode on, and make sure you know specifically what you are testing to watch for. We originally didn't think about this, it was something pretty simple to miss. We had the "query" output, but didn't think that the query may be cached, for example, so sometimes it may not have been printed, but did execute on a given server node and we thought it wasn't working. Anyway, myself and another chap in the forums are considering writing initially a simple, but eventually a generic comprehensive J2ee clustering test application where by various scenarios would be handled to ensure clustering is working, and also provide some sort of performance specs and graphing as to a given setups scalability and so forth. Before we venture on to this, is there any free or open-source application that is already in progress like this? Thanks. __________________________________ Do you Yahoo!? SBC Yahoo! DSL - Now only $29.95 per month! http://sbc.yahoo.com ------------------------------------------------------- This SF.Net email sponsored by: Parasoft Error proof Web apps, automate testing & more. Download & eval WebKing and get a free book. www.parasoft.com/bulletproofapps1 _______________________________________________ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
