Laszlo Gaal created IMPALA-13862:
------------------------------------
Summary: test_int64_timestamps fails in Dockerised tests running
on Rocky 8.5
Key: IMPALA-13862
URL: https://issues.apache.org/jira/browse/IMPALA-13862
Project: IMPALA
Issue Type: Bug
Components: Backend
Affects Versions: Impala 5.0.0
Reporter: Laszlo Gaal
When dockerised EE_TESTs are executed on Rocky 8.5, test_int64_timestamp
exhibits a single failed test. It is repeated 8 times as the test is run
through different test vectors, but all 8 failures are actually the same item.
The stack trace:
{code}
query_test/test_scanners.py:1172: in test_int64_timestamps
'QueryTest/parquet-int64-timestamps', vector, unique_database)
common/impala_test_suite.py:904: in run_test_case
self.__verify_results_and_errors(vector, test_section, result, use_db)
common/impala_test_suite.py:737: in __verify_results_and_errors
replace_filenames_with_placeholder)
common/test_result_verifier.py:523: in verify_raw_results
VERIFIER_MAP[verifier](expected, actual)
common/test_result_verifier.py:305: in verify_query_result_is_equal
assert expected_results == actual_results
E assert Comparing QueryTestResults (expected vs actual):
E -2147483648,1970-01-01 00:59:57.852516352,1969-12-31 23:59:57.852516352
== -2147483648,1970-01-01 00:59:57.852516352,1969-12-31 23:59:57.852516352
E -9223372036854775808,1677-09-21 01:12:43.145224192,1677-09-21
00:12:43.145224192 != -9223372036854775808,1677-09-21
00:30:13.145224192,1677-09-21 00:12:43.145224192
E 0,1970-01-01 01:00:00,1970-01-01 00:00:00 == 0,1970-01-01
01:00:00,1970-01-01 00:00:00
E 2147483647,1970-01-01 01:00:02.147483647,1970-01-01 00:00:02.147483647 ==
2147483647,1970-01-01 01:00:02.147483647,1970-01-01 00:00:02.147483647
E 9223372036854775807,2262-04-12 01:47:16.854775807,2262-04-11
23:47:16.854775807 == 9223372036854775807,2262-04-12
01:47:16.854775807,2262-04-11 23:47:16.854775807
E Detailed information truncated (-4 more lines), use "-vv" to show{code}
(The blank lines around the failing comparison were added for clarity, they
were not part of the test output.)
This is seen on the master branch.
The failure looks OS-specific, or specific to Red Hat derivatives: dockerised
tests on Ubuntu 20.04 are parts of our precommit process and they have not
exhibited this failure. The symptoms were first seen during development for
IMPALA-13825 on Rocky 9.2, then reproduced on private infrastructure on Red Hat
9.2; this led to the repro scenario on Rocky 8.5.
The only successful run retained on jenkins.impala.io for this configuration is
from August 2023, using commit {{75bf21341800aaef87b0fb85af09417fd0a34820}}
Dockerised tests run the minicluster just like in regular builds (processes
running natively on the host), but all the Impala roles are run in individual
Docker containers. The container build logic is located in
${IMPALA_HOME}/docker.
The base OS image for these containers is chosen dynamically to match the
host's OS, so that impalad binaries can operate both inside and outside the
containers.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
