Details
-
Bug
-
Status: Resolved
-
Blocker
-
Resolution: Fixed
-
Impala 2.6.0
Description
Failure first showed up in 5.8.0-non-partitioned-joins-and-aggs build:
16:41:52 TestTPCHJoinQueries.test_outer_joins[batch_size: 0 | exec_option: {'disable_codegen': False, 'abort_on_error': 1, 'exec_single_node_rows_threshold': 0, 'batch_size': 0, 'num_nodes': 0} | table_format: parquet/none]
16:41:52 [gw3] linux2 -- Python 2.6.6 /data/jenkins/workspace/impala-umbrella-build-and-test/repos/Impala/bin/../infra/python/env/bin/python
16:41:52 query_test/test_join_queries.py:97: in test_outer_joins
16:41:52 self.run_test_case('tpch-outer-joins', new_vector)
16:41:52 common/impala_test_suite.py:288: in run_test_case
16:41:52 assert expected_str in actual_str
16:41:52 E assert 'Repartitioning did not reduce the size of a spilled partition' in 'ImpalaBeeswaxException: Query aborted:Memory limit exceededMemory Limit ExceededQuery(bc43579b6559def6:a6e1fb31b15377... EXCHANGE_NODE (id=5): Consumption=0 DataStreamRecvr: Consumption=5.41 MB DataStreamSender: Consumption=4.00 KB'
http://sandbox.jenkins.cloudera.com/job/impala-umbrella-build-and-test/871/
http://sandbox.jenkins.cloudera.com/job/impala-cdh5-2.6.0_5.8.0-core-non-partitioned-joins-and-aggs/3/
sailesh marked this build as keep forever. Should unmark it once this is fixed.
The test error message was recently updated as the error message was updated in PHJ when repartitioning didn't reduce the size of a partition. This apparently doesn't jive well with the non-partitioned hash joins and occasionally, the test can still fail with "memory limit exceeded" error message even in partitioned hash join:
http://sandbox.jenkins.cloudera.com/view/Impala/view/Evergreen-cdh5-trunk/job/impala-cdh5-trunk-core-local-filesystem/62/console
http://sandbox.jenkins.cloudera.com/view/Impala/view/Evergreen-cdh5-2.6.0_5.8.x/job/impala-cdh5-2.6.0_5.8.0-exhaustive-release/4/