4 Replies Latest reply on Mar 29, 2010 10:06 AM by adsoko

    Jboss cluster being restarted by thin user

      We are running   jboss-cloud/1.0.0.Beta8-SNAPSHOT

      in /var/log/jboss-cloud-management/default.log

       

      there is this entry:

       


      D, [2010-03-29T07:19:03.183044 #1790] DEBUG -- : Handling event management_address_request for prefix latest...
      D, [2010-03-29T07:19:03.183241 #1790] DEBUG -- : Notyfing handler JBossCloudManagement::ManagementAddressRequestHandler
      I, [2010-03-29T07:19:03.183317 #1790]  INFO -- : Got new management-appliance address: 10.226.234.239
      I, [2010-03-29T07:19:03.183388 #1790]  INFO -- : Asking for front-end appliance address...
      I, [2010-03-29T07:19:03.244498 #1790]  INFO -- : Injecting front-and appliance address 10.227.94.83 to '/etc/jboss-as5.conf'...
      I, [2010-03-29T07:19:03.362601 #1790]  INFO -- : Stopping jboss-as5 service...
      D, [2010-03-29T07:19:03.363484 #1790] DEBUG -- : Event management_address_request was successfuly handled

       

      ...

       


      D, [2010-03-29T07:21:31.760384 #1790] DEBUG -- : Notyfing handler JBossCloudManagement::ManagementAddressRequestHandler
      I, [2010-03-29T07:21:31.760472 #1790]  INFO -- : Got new management-appliance address: 10.226.234.239
      I, [2010-03-29T07:21:31.760558 #1790]  INFO -- : Asking for front-end appliance address...
      D, [2010-03-29T07:21:31.767066 #1790] DEBUG -- : Event management_address_request was successfuly handled
      I, [2010-03-29T07:21:38.734311 #1790]  INFO -- : Service jboss-as5 successfully stopped.
      I, [2010-03-29T07:21:38.734514 #1790]  INFO -- : Starting jboss-as5 service...

      don't understand why the front-end address is changing suddenly.
      Is this a bug that has been fixed in later versions ?

      Many thanks
        • 1. Re: Jboss cluster being restarted by thin user
          goldmann

          Hi Admir,

           

          In JBoss Cloud when the front-end address is changing, new address is being injected to all back-end's. This causes a restart of back-end to activate the changes. There could be a two issues:

          1. management service running on back-end appliance simply "forgets" the front-end appliance address (obviously a bug),
          2. front-end was really crashed and restarted with a new IP address (normal behaviour),

           

          From CirrAS 1.0.0.Beta2 we don't require a restart of JBoss AS after front-end change. New front-end addresses are injected now using JMX calls. No matter what happens, your back-end will be still running.

           

          --Marek

          • 2. Re: Jboss cluster being restarted by thin user

            Hi Marek

             

            We can't upgrade to the latest CirrAS because it runs on jboss 6. For various reasons we are still at jboss 5.

             

            The 'change of address' message is received simultaneously by all back-end appliances. So the message is genuine.

            If I would check the front end appliance or the management appliance for the source of the problem, where would I look for any clues ?

             

            Thanks

            Admir

            • 3. Re: Jboss cluster being restarted by thin user
              goldmann
              We can't upgrade to the latest CirrAS because it runs on jboss 6. For various reasons we are still at jboss 5.

              Yes, this could be an issue. But still IMO JBoss AS 5 is easy upgradeable to AS 6.

              The 'change of address' message is received simultaneously by all back-end appliances. So the message is genuine.

              If I would check the front end appliance or the management appliance for the source of the problem, where would I look for any clues ?

              I would check if front-end address was really changed, so scan default.log:

              grep "Asking for front-end appliance address" -A 1

              When the address is same as previous it shouldn't be injected into /etc/jboss-as5.conf. If it is injected the next step is to fix this in /usr/share/jboss-cloud-management/lib/jboss-cloud/management/api/2009-05-18/handler/put/management-address-request-handler.rb. Take a look at line 79 of this file. If this is a bug, there is the good place to fix it.

               

              Are you running more than 1 front-end appliance? This could be also the cause...

               

              HTH

               

              --Marek

              • 4. Re: Jboss cluster being restarted by thin user

                The front end address changed in the back end appliances. found: Asking for front-end appliance address...

                This was found in the back-end boxes but not in the front-end boxes.

                 

                 

                But on the other hand we have two front end appliances. Are we doing something wrong in their configuration ? or would they cause this problem regardless.