[IMPALA-5198] Error messages are sometimes dropped before reaching client

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: Impala 2.8.0
• Fix Version/s: Impala 2.9.0
• Component/s: Infrastructure
• Labels: broken-build
• Epic Color: ghx-label-6
• Assignee: Sailesh Mukil
• Reporter: Matthew Jacobs
• Created: 11/Apr/17 21:28
• Updated: 21/Apr/17 03:41
• Resolved: 21/Apr/17 03:41

Description

On the nightly core s3 tests, the custom cluster test test_exchange_delays failed. It looks like the query failed for an unexplained reason. It says 'aborted' but we need to look at it more closely.

07:56:58 =================================== FAILURES ===================================
07:56:58  TestExchangeDelays.test_exchange_small_delay[exec_option: {'disable_codegen': False, 'abort_on_error': 1, 'exec_single_node_rows_threshold': 0, 'batch_size': 0, 'num_nodes': 0} | table_format: text/none] 
07:56:58 
07:56:58 self = <test_exchange_delays.TestExchangeDelays object at 0x3f054d0>
07:56:58 vector = <tests.common.test_vector.ImpalaTestVector object at 0x5a32750>
07:56:58 
07:56:58     @pytest.mark.execute_serially
07:56:58     @CustomClusterTestSuite.with_args("--stress_datastream_recvr_delay_ms=10000"
07:56:58           " --datastream_sender_timeout_ms=5000")
07:56:58     def test_exchange_small_delay(self, vector):
07:56:58       """Test delays in registering data stream receivers where the first one or two
07:56:58         batches will time out before the receiver registers, but subsequent batches will
07:56:58         arrive after the receiver registers. Before IMPALA-2987, this scenario resulted in
07:56:58         incorrect results.
07:56:58         """
07:56:58 >     self.run_test_case('QueryTest/exchange-delays', vector)
07:56:58 
07:56:58 custom_cluster/test_exchange_delays.py:39: 
07:56:58 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
07:56:58 common/impala_test_suite.py:362: in run_test_case
07:56:58     self.__verify_exceptions(test_section['CATCH'], str(e), use_db)
07:56:58 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
07:56:58 
07:56:58 self = <test_exchange_delays.TestExchangeDelays object at 0x3f054d0>
07:56:58 expected_strs = ['Sender timed out waiting for receiver fragment instance\n']
07:56:58 actual_str = 'ImpalaBeeswaxException: Query aborted: (1 of 3 similar)'
07:56:58 use_db = None
07:56:58 
07:56:58     def __verify_exceptions(self, expected_strs, actual_str, use_db):
07:56:58       """
07:56:58         Verifies that at least one of the strings in 'expected_str' is a substring of the
07:56:58         actual exception string 'actual_str'.
07:56:58         """
07:56:58       actual_str = actual_str.replace('\n', '')
07:56:58       for expected_str in expected_strs:
07:56:58         # In error messages, some paths are always qualified and some are not.
07:56:58         # So, allow both $NAMENODE and $FILESYSTEM_PREFIX to be used in CATCH.
07:56:58         expected_str = expected_str.strip() \
07:56:58             .replace('$FILESYSTEM_PREFIX', FILESYSTEM_PREFIX) \
07:56:58             .replace('$NAMENODE', NAMENODE) \
07:56:58             .replace('$IMPALA_HOME', IMPALA_HOME)
07:56:58         if use_db: expected_str = expected_str.replace('$DATABASE', use_db)
07:56:58         # Strip newlines so we can split error message into multiple lines
07:56:58         expected_str = expected_str.replace('\n', '')
07:56:58         if expected_str in actual_str: return
07:56:58       assert False, 'Unexpected exception string. Expected: %s\nNot found in actual: %s' % \
07:56:58 >       (expected_str, actual_str)
07:56:58 E     AssertionError: Unexpected exception string. Expected: Sender timed out waiting for receiver fragment instance
07:56:58 E     Not found in actual: ImpalaBeeswaxException: Query aborted: (1 of 3 similar)
07:56:58 
07:56:58 common/impala_test_suite.py:253: AssertionError

07:56:58 ----------------------------- Captured stderr call -----------------------------
07:56:58 -- executing against localhost:21000
07:56:58 use functional;
07:56:58 
07:56:58 SET disable_codegen=False;
07:56:58 SET abort_on_error=1;
07:56:58 SET exec_single_node_rows_threshold=0;
07:56:58 SET batch_size=0;
07:56:58 SET num_nodes=0;
07:56:58 -- executing against localhost:21000
07:56:58 select count(*)
07:56:58 from tpch.lineitem
07:56:58   inner join tpch.orders on l_orderkey = o_orderkey;
07:56:58 
07:56:58 ======== 1 failed, 43 passed, 20 skipped, 8 xfailed in 3213.63 seconds =========

Issue Links

• relates to: IMPALA-4697 RuntimeState::LogError() does not log merged status