0 Replies Latest reply on Sep 2, 2006 10:02 AM by sdsinc_pmascari

    CFMX7 Clusters and Load Balancing

    sdsinc_pmascari Level 2
      I have CFMX 7 set up in a distributed cluster: 2 Windows 2003/IIS6 servers with Network Load Balancing serving 2 CFMX7 Enterprise Servers (on Win2K) each set up with 4 identical CF instances. Each instance is clustered with it's corresponding instance on the other CF server.

      We have set up this way for redundancy and load balancing sake.

      When we first set this up in a development environment it worked great. But we've since moved it to a production environment and, in doing so, had to change the server's IP addresses and, thus, we needed to "rebuild" the clusters. This all went off without a hitch and looking through the CF administrator the instances are all running normally and the clusters are configured just fine. However, the load balancing and redundancy features no longer work! We can browse our sites without issue, but they always open from CF Server 1 and never from the clustered CF Server 2 even though the cluster algorith is set to Round Robin. Also, if I shut down a CF instance on CF Server 1 I would expect the request be re-routed to CF Server 2, as it did in our development environment, but it no longer does this. With CF Server 1 shut down, whe I try to pull up our site I get "waiting for mysite.com" in my browser status bar and the site never loads. I then start CF Server 1 back up, reload the page and we're back in business.

      Lastly, to emphasize that all requests are going to only one server, we have a Google Mini box that indexes our sites periodically. If we start that machine crawling and watch Windows Task Manager we see the CF Server 1 is getting all the hits while CF Server 2 appears to be getting none...just sitting there with System Idle Process at 99%

      What gives? Why would this have stopped working?

      One other thing, when I changed the IP addresses of the servers and re-built the instances, I gave the instances the same names that I had given them in development. Is it possible there is some kind of "named-based" config file that wasn't updated properly and it is now "invisibly" failing to cluster because it's pointing to the wrong IP?

      Looking for any ideas here...