Spark
As a general-purpose data engine, Apache Spark can integrate with Hive closely. Spark SQL has supported a subset of HQL and can leverage the Hive metastore
to write or query data in Hive. This approach is also called Spark over Hive. To configure Spark, use Hive the metastore
, you only need to copy the hive-site.xml
to the ${SPARK_HOME}/conf
directory. After that, running the spark-sql
command will enter the Spark SQL interactive environment, where you can write SQL to query Hive tables.
On the other hand, Hive over Spark is a similar approach, but lets Hive use Spark as an alternative engine. In this case, users still stay in Hive and write HQL, but run over the Spark engine transparently. Hive over Spark requires the Yarn FairScheduler
andset hive.execution.engine=spark
. For more details, refer to https://cwiki.apache.org/confluence/display/Hive/Hive+on+Spark%3A+Getting+Started.