The Hive Storage Handlers framework is used for the Teradata QueryGrid Hive initiator connector. Storage handlers allow Hive to access data stored on other systems. A Hive table created using a storage handler is referred to as a non-native table.
Hive storage handlers operate at the table level, so one non-native table must be created for each remote table, unlike Teradata Foreign Servers and Presto Catalogs that operate at the database level.
The following steps provide an example of configuring a non-native table in order to use it with a Hive-to-Teradata link.
- Set the link properties for the Hive-to-Teradata link in the QueryGrid portlet.
- Log on to a Hive client such as Beeline.
- Add hive-loaderfactory JAR to the class path:add jar /opt/teradata/tdqg/connector/tdqg-hive-connector/version/lib/hive-loaderfactory-version.jar;
There are others ways to add JARs to Hive class path, one of which is by setting HIVE_AUX_JARS_PATH.
The last approach typically requires a Hive restart. If you are using a Cloudera implementation of Hadoop and Sentry is enabled, this may be the only option available for use because Sentry does not allow use of ADD JAR commands. - Use a Hive storage handler to create the non-native table. For example:
CREATE EXTERNAL TABLE cardata_remote ROW FORMAT SERDE 'com.teradata.querygrid.qgc.hive.QGSerDe' STORED BY'com.teradata.querygrid.qgc.hive.QGStorageHandler' TBLPROPERTIES ( "link"="hive_to_td_link", "version"="active", "table"="ut1.cardata");
There is no column definition in the CREATE TABLE statement because the column definition is inferred dynamically by the Serializer and Deserializer (SerDe).The table property clause contains the fully-qualified name of the remote table; for example, database.schema.table, schema.table, and table. If any part is omitted, default values (specified in connector properties) are used.