Joe McDonnell created IMPALA-9266:
-------------------------------------
Summary: TestLogFragments.test_log_fragments fails due to missing
log
Key: IMPALA-9266
URL: https://issues.apache.org/jira/browse/IMPALA-9266
Project: IMPALA
Issue Type: Bug
Components: Infrastructure
Affects Versions: Impala 3.4.0
Reporter: Joe McDonnell
TestLogFragments.test_log_fragments is failing due to missing a log entry:
{noformat}
/data/jenkins/workspace/impala-asf-master-core/repos/Impala/tests/observability/test_log_fragments.py:46:
in test_log_fragments
"] Analysis and authorization finished.")
common/impala_test_suite.py:1149: in assert_impalad_log_contains
self.assert_log_contains("impalad", level, line_regex, expected_count)
common/impala_test_suite.py:1185: in assert_log_contains
(expected_count, log_file_path, line_regex, found, line)
E AssertionError: Expected 1 lines in file
/data0/jenkins/workspace/impala-asf-master-core/repos/Impala/logs/ee_tests/impalad.impala-ec2-centos74-m5-4xlarge-ondemand-088c.vpc.cloudera.com.jenkins.log.INFO.20191227-001949.23945
matching regex 'ce41d657e70d6890:6f0f227d00000000] Analysis and authorization
finished.', but found 0 lines. Last line was:
E Caught signal: SIGTERM. Daemon will exit.{noformat}
This started happening after the "IMPALA-8974: Fixed a bug when create kudu
managed table without HMS" commit went in. That commit adds a test that
restarts Impala in a frontend test. The problem is that it runs
start-impala-cluster.py without arguments, whereas bin/run-all-tests.sh runs
start-impala-cluster.py specifying the --log_dir. This would put the log files
in a different location (/tmp?).
[https://github.com/apache/impala/blob/320f05852060c1027326ac20be7df340a7a5263f/fe/src/test/java/org/apache/impala/catalog/CreateKuduTableWithoutHMSTest.java#L98]
[https://github.com/apache/impala/blob/master/bin/run-all-tests.sh#L165-L167]
In one run that hit this issue, there are two sets of impalad logs in the
ee_test directory. One set starts at 06:40:22 and ends at 07:11:28. The second
set starts at 09:45:25 and ends at 09:47:30. So, this is missing 2.5 hours of
ee_test log files, which matches the theory.
This is also likely to impact other things like erasure coding or tests that
run against the data cache.
GVO doesn't hit this because the job that runs frontend tests does not run end
to end tests.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]