Spark in addition to the source release provides several binary distributions (https://spark.apache.org/downloads.html).
1. Hadoop and Hive included. It is the version that Spark was built and tested with. 2. User provided Hadoop and Hive. Allows users to provide their own Hadoop and Hive libraries. I see that tez-api.jar is part of the Hive distribution. When user provides a different versions of the Tez, won’t that cause a problem as now there will be two different versions on the classpath? Thank you, Vlad On Jun 19, 2025, at 7:56 AM, Denys Kuzmenko <dkuzme...@apache.org> wrote: Hi Vlad, It's documented in https://hive.apache.org/docs/latest/manual-installation_283118363/. I believe that was done to avoid tight coupling, allowing different versions of the Tez runtime to be used. Regards, Denys
