We have an CDH 5.8.3 cluster. We use tableau 10.2.2 with the latest Impala ODBC driver.
After some hours of work the Tableau datasource, which is one node of the cluster stops accepting new connections.
If I point to another node it will work normally, and also the original node which I was referring to will execute query fragments.
It looks like we are filling up something on hiveserver2 part of impalad. I tried to increase fe_service_threads to 512 but it didn't solve the issue.
Do I need to increase also the be_service_threads? I understood that it is useless.
Is there anything I can look at to understand what limit I'm touching?