Many users and BI tools use JDBC connections in order to retrieve data. As Livy exposes only a REST API, this is a limitation in its adoption. Hence, adding a JDBC endpoint may be a very useful feature, which could also make Livy a more attractive solution for end user to adopt.
Moreover, currently, Spark exposes a JDBC interface, but this has many limitations, including that all the queries are submitted to the same application, therefore there is no isolation/security, which can be offered by Livy, making a Livy JDBC API a better solution for companies/users who want to use Spark in order to run they queries through JDBC.
In order to make the transition from existing solutions to the new JDBC server seamless, the proposal is to use the Hive thrift-server and extend it as it was done by the STS.
|Dedicated thread for timeout checker||Open||Unassigned|