Github user mgaido91 commented on the issue: https://github.com/apache/spark/pull/20029 What it seems is never closed by your analysis is the client used to interact with the metastore. This might be a problem which we are not aware of in normal SQL applications, since we have only one client in those cases. What you are doing in your fix is avoiding creating a client for each `HiveSessionBuilder`, thus: 1. this would mean that we are creating more than one `SessionBuilder`, ie. more than one `SparkSession`, which is not true as far as I know. 2. any session would share the same client to connect to the metastore, which is wrong IMHO. Please let me know if I misunderstood or I was wrong with something.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org