I have a Hortonworks Sandox running HDP 2.4 on a Oracle Virtual Box with 15 GB RAM assigned on a CentOS 6 Dual Core machine. I have connected my Tableau report from another machine to the Spark in the HDP. My sample report has close to 120 tables with a Star and Snowflake schema design. All the FACT tables are empty as of now. When I open this report in Tableau 10.1 the SparkUI, first creates 3 jobs. Each job has 210 stages. Also, each job has 21,000 tasks.
Now comes the problem - After 70 mins of report execution, and NO result, the SparkUI creates a 4th job, again with 21K tasks and after 10 minutes, a 5th job is created, again with 21K tasks. It seems like to be a never ending process. The RAM, which was initially 8 GB has come down to 150 MB but, both, SparkUI and Tableau reports keep on executing. Not sure what is happening here. I have tried connecting to Spark server from a new tableau report and that is also working fine.