3 Replies Latest reply on Jan 10, 2019 1:30 AM by Richard Silva

    Question on Extract

    Madhusudhan Khambham

      Hi Experts,

       

      I have a Primary Server, 6 worker machines with 2 Backgrounder instances in each worker.

      Now, when the Backgrounder of worker A ran an extract successfully.

       

      Here are the questions i have.

       

      1. This will save the data on the Worker/Backgrounder that the extract ran, correct?

      2. How it will share it with the other 5 worker machines?

      3. What if the Worker A machine got crashed?

      4. in this case how it will share the data to all the other workers?

      5. will it affect the reports that are using the extracted data using Worker A(when it is in crashed state)?

      6. When the extract is successful in Worker A, will it sync the data to all the remaining workers?

      7. If yes to point # 6, then we are saving the data 6 times(as i have 6 workers). This will be redundant correct?

       

      Appreciated your time in advance for answering the questions.

       

      Thanks & Regards

      Madhu K

        • 1. Re: Question on Extract
          Jeff Strauss

          General info:

          It will save the extract onto whatever nodes you have the dataengine process active.  It replicates the created extract via the filestore process which also keeps the data in sync across the distributed dataengine processes.  If a node only has the vizql process and no dataengine process, then Tableau will route traffic to one of the other nodes where the dataengine is active in order to retrieve the required info.

           

          1. It will save onto the node where the backgrounder processed only if there is an active dataengine process

          2. It will share internally within the cluster dependent on where the dataengine process(es) live.  Tableau knows how to communicate across the cluster via the cluster controller

          3. If you have the dataengine process allocated on other workers, then you should be ok as the filestore process has replicated the data and kept it in sync.  See High Availability for more info

          4. Either vizql can talk with data extracts on whatever node they live on, or the filestore replicates it to the specific nodes

          5. Not sure what you mean

          6. Yes, if the dataengine is allocated on other workers

          7. Yes

          2 of 2 people found this helpful
          • 2. Re: Question on Extract
            Madhusudhan Khambham

            Thank you for the response Jeff, This cleared some of my doubts.

            • 3. Re: Question on Extract
              Richard Silva

              Hi Jeff Strauss,

               

               

              Thank you for the information that you have shared. Just a follow-up question regarding this. Please note that we currently have 4 worker nodes.

               

              1. You have said that upon successful extract in Worker A, it will sync all the data to all remaining workers and it will be redundant across all workers. With that said, can we reduce the # of worker nodes without affecting the performance of the server? I am looking for ways to help our company save some money in infrastructure to be able to allot more money in buying server licenses.

              2. Can you provide a recommendation for the settings/configuration of the Tableau Server with 4 worker nodes? (I know there is a documentation regarding this but I can't fully understand it. Sorry)

               

              Hoping for a favorable response on this one. Thank you.

               

              Regards,

              Richard Silva