6 Replies Latest reply on Apr 7, 2017 10:02 AM by Soundararajan Ananthu Ramalingam

    Worker node configuration

    Soundararajan Ananthu Ramalingam

      We have installed the tableau primary server and two worker nodes.  The worker nodes are configured with the primary ip address.

       

      In case the primary server failed and not able to restore.

       

      If we provisioning the new primary server, then how to reconfigure the worker node with the new primary server IP address?

       

      What is the best practice to avoid the reconfiguration other than hostname or ip address?

        • 1. Re: Worker node configuration
          Glen Robinson

          Hi Soundararajan

          The Tableau best practice is to create a Backup Primary Server

          (Please read the following)

          Use a Backup Primary

           

          This is when you have a High Availability Cluster, ( where the Repository is not running on the Primary Server).

          You create another Server running Tableau (call the Backup Primary Server), and configure the cluster to be aware of this additional server.

          In the event of the Primary server becoming unavailable, you can then bring the Backup Primary online, and introduce it to the cluster.

           

          Alternatively, if you are running your Primary server on a virtual platform and are able to take nightly image backups of the primary server, then in the event of the primary server becoming unavailable, you can revert to the previous night's backup.

          Hope this helps

          Glen

          • 2. Re: Worker node configuration
            Soundararajan Ananthu Ramalingam

            Hi Glen,

             

            Sorry I am looking for the solution where the worker node is configured with the IP address of the primary server.

             

            If we provision the new primary server, when you add the worker node in the primary server, it will not recognize the IP address.  The only option we have is to uninstall the worker node software and re-install with the new IP address.

             

            In case the worker node is configured with the hostname, then worker node restart will be sufficient.

             

            Do we have any solution to avoid this situation of re-install in case of IP address configuration?

             

            Thanks for sharing the backup primary, the solution to make backup primary can be automated through script I guess, there is no configuration level parameter.  Please correct me I am wrong.

             

            Thanks

            Soundar

            • 3. Re: Worker node configuration
              Glen Robinson

              Hi

              Just to clarify.

              You are after a solution for the following?

              1. Primary Server Dies.

              2. You build a brand new server, with a new Server name / IP Address, and install Tableau on it

              3. You want to be able to tell your workers that this is the new Primary without reinstallation...

              4. The IP address of the 'new' Server could be anything, so we are unable to pre-define this on the workers right now.

              Is this right?

               

              For the backup primary option, failover to the backup primary can be automated through a script.

              There is downtime as the tableau cluster needs to be stopped and started to add the backup primary server.

              The Worker server do need to be configured prior to a failover happening, so that they are aware that the backup primary is there. (using the tabadmin failoverprimary command)

               

              Glen

              • 4. Re: Worker node configuration
                Soundararajan Ananthu Ramalingam

                Hi Glen,

                You are after a solution for the following?

                1. Primary Server Dies.

                2. You build a brand new server, with a new Server name / IP Address, and install Tableau on it

                3. You want to be able to tell your workers that this is the new Primary without reinstallation...

                4. The IP address of the 'new' Server could be anything, so we are unable to pre-define this on the workers right now.

                Is this right?  Yes, I am looking for the best solution here. 

                 

                Is there any best practice on the hostname on the primary and worker nodes (like tableau_server, tableau_worker1, tableau_worker2)?

                 

                There is downtime as the tableau cluster needs to be stopped and started to add the backup primary server.

                The Worker server do need to be configured prior to a failover happening, so that they are aware that the backup primary is there. (using the tabadmin failoverprimary command)

                 

                Is the command tabadmin failoverprimary can be automated?

                 

                Thanks

                Soundar

                • 5. Re: Worker node configuration
                  Glen Robinson

                  Hi Soundararajan

                  I'm not sure that there is best solution here.

                  I would say that your options are as follows

                  1. Follow the official Tableau route of setting up a Backup Primary Server

                  2. Use Virtual Image backups (if possible)

                  3. If you need to build a new server, use the same name / IP address as the original Primary Server.

                   

                  I am sure that I have tried (and failed) doing what you are attempting. (trying to add a new primary server into a cluster)

                   

                  AFAIK, The tabadmin failover command initially needs to be run whilst the cluster is working, so that the working cluster is aware that there is a backup primary.

                  When the primary server has failed and you are promoting the backup primary, the command is run again to promote the backup primary. (This step of process can be automated / batched)

                   

                  Regards

                  Glen

                  1 of 1 people found this helpful
                  • 6. Re: Worker node configuration
                    Soundararajan Ananthu Ramalingam

                    Thank you Glen.  We can close this thread.  

                    I appreciate you response on my other question "Site specific performance and monitoring"