Hi, All. SPARK-22279 <https://issues.apache.org/jira/browse/SPARK-22279> turned on `convertMetastoreOrc` by default for `Feature Parity with Parquet`.
Unfortunately, it will be turned off back via https://github.com/apache/spark/pull/20536 in Apache Spark 2.3 RC3 because a well-known `convertMetastoreOrc/Parquet` issue is not resolved yet. That means reading Hive ORC tables is not accelerated by vectorization in a vanilla configuration. This will prevent any regression on Hive ORC tables and users are still able to turn on it manually. Sorry for last minute notification and thank you for your support and decision. Bests, Dongjoon.