Hive

Bug Fixes

  • HIVE-3032: Fixes a problem that caused INSERT OVERWRITEs to fail when hive.allow.move.on.s3 was set to true and the table name contained = . This problem occurred when the table location path was longer or shorter than the partition path generated (dynamically when hive.allow.move.on.s3 was set to true) and resulted in invalid partition specs.
  • HIVE-3035: Fixes a problem that causes a dynamic write to a partitioned table to fail if a partition already exists with a custom location. The fix consists of skipping the exception block if hive.qubole.dynpart.move.check is set to false.
  • HIVE-3041: Fix for proper handling of Thrift metastore exceptions which were causing GRANT, CREATE, and DROP commands to hang (in Hive 2.1.1) or fail with a null error (in Hive 1.2) when the object already existed in the metastore.
  • HIVE-3042: Fixes a problem that caused a Hive query to hang during job submission.
  • HIVE-3133: Fix to handle empty vector column batches in the FLOAT and DOUBLE reader when you use the Vectorized Reader.
  • HIVE-3166: Fixes a problem that occurred in reading a Hive bootstrap script.
  • HIVE-3172: Fixes a NPE (Null Pointer Exception) in the Historical SQL Operations section of the HiveServer2 web UI.

Enhancements

  • HIVE-3030: Tez can now handle a scenario where a single key-value pair size does not fit into any of the buffer blocks. This issue caused some Hive queries to fail with BufferOverFlowError when Tez was used.
  • HIVE-3167: Fixes a race condition which caused queries to fail with State Transition Exception. Related Open Source JIRAs: HIVE-15563, HIVE-17352.