3 Replies Latest reply on Apr 14, 2011 7:03 AM by rhusar

    Service Unavailable Error

    massios
    massios Apr 14, 2011 6:29 AM

      Hello mod_cluster community,

       

      We have two physical servers running Microsoft Windows Server 2003. Each windows machine is running an Apache http 2.2.17 with mod_cluster 1.1. Further, each physical server is running two JBoss AS 5.1 GA nodes with a JBoss ESB installed. In total we have 4 jboss AS nodes in a cluster configuration. The mod_cluster plugin is loadbalancing http requests along the 4 jboss nodes of the cluster. There is a hardware loadbalancer to loadbalance requests on the two Apache http 2.2.17 servers.

       

      We are running in this configuration for about 2 months now without any problems until yesterday. Yesterday all of a sudden one of the two apache http servers started reporting "Service Unavailable" errors. The 4 jboss as nodes behind the apache node that was failing seemed to be ok and were replying to requests either directly or through the second apache server.

       

      The log of the apache http looks like the attached file

       

      Do you have any clues that you can give us for what to look for? The problem was solved after a complete restart of jboss and apache but this is not a nice solution.

       

       

      Nikos

      • 4012 Views
      • Tags:
        • 1. Service Unavailable Error
          rhusar
          rhusar Apr 14, 2011 6:38 AM (in response to massios)

          Hi Nikos, so are you saying  that restarting Apache HTTPd was not enough and you had to restart all JBoss instances as well?

            • Actions
          • 2. Service Unavailable Error
            massios
            massios Apr 14, 2011 6:54 AM (in response to rhusar)

            Hello Radoslav,

             

            Unfortunately yes

            .

            The operators have restarted the failed apache http several times during the day. What was strange was that although one of the http servers was stuck the other apache  was distributing requests to all 4 jboss nodes without any problems.

             

            We had in the end to restart the jboss instances as well. We had to restart all the servers in the cluster. 4 jboss instances and 2 apaches. Just Restarting  the servers on the one physical server was not enough.

             

            Nikos

              • Actions
            • 3. Service Unavailable Error
              rhusar
              rhusar Apr 14, 2011 7:03 AM (in response to massios)

              Hi Nikos, sad thing. I havent seen that myself yet - restarting should often help as long as the mod_cluster.sar part sends information to the Apache. Probably the best idea would be to switch to 1.1.1, I havent checked precisely the log but some related critical issues have been fixed there, have a look please: http://docs.jboss.org/mod_cluster/1.1.0/html/changelog.html#d0e3234

                • Actions