Improper Load BalancingI can't give you a solid answer. I'm currently working on the same sort of issue (in my spare time), so it's just my observations of what I've looked at so far. We have windows servers running Weblogic to Oracle Rac on linux. I noticed most of the connections going to one server one day, another the next. This makes sense as the windows servers are rebooted often and connection pooling is used. So if there's no load, the pools would all go to one instance. But there are some apps that connect and leave with no pooling. Clients connnection string has load_balance=yes. I've only been on RAC for 5 weeks now so I'm a little green I ran a series of tests from multiple clients, basically opening up a few hundred connections and seeing where they connect. I would modify the connection string and at one point, hardcoded a specific node ip into the connect string. The connections still went to a different node. My conclusion was that the listener on the rac was diverting connections which is fair enough I suppose. So, on my machines "lsnrctl services" doesn't return anything while "lsnrctl status" does. I believe that the instances are not registering automatically with the listener as none of the listener parameters are set in the init.ora file. I would like to do some more testing but my only RAC is prod and SLA's are very tight.
I followed note 226880.1 on metalink. -- //www.freelists.org/webpage/oracle-l
