Details
-
Bug
-
Status: Resolved
-
Critical
-
Resolution: Fixed
-
Impala 2.9.0
Description
Found by David Knupp: TestCancellationSerial.test_cancel_insert is so flaky that it doesn't even run the same number of test cases each time:
jbapple@jbapple-optiplex:~/Impala/tests:((detached from gerrit-asf/master))$ impala-py.test --exploration_strategy=core --collect-only query_test/test_cancellation.py | grep -c insert 7 jbapple@jbapple-optiplex:~/Impala/tests:((detached from gerrit-asf/master))$ impala-py.test --exploration_strategy=core --collect-only query_test/test_cancellation.py | grep -c insert 4 jbapple@jbapple-optiplex:~/Impala/tests:((detached from gerrit-asf/master))$ impala-py.test --exploration_strategy=core --collect-only query_test/test_cancellation.py | grep -c insert 3 jbapple@jbapple-optiplex:~/Impala/tests:((detached from gerrit-asf/master))$ impala-py.test --exploration_strategy=core --collect-only query_test/test_cancellation.py | grep -c insert 8 jbapple@jbapple-optiplex:~/Impala/tests:((detached from gerrit-asf/master))$ impala-py.test --exploration_strategy=core --collect-only query_test/test_cancellation.py | grep -c insert 7
http://jenkins.impala.io:8080/job/ubuntu-14.04-from-scratch/657 and http://jenkins.impala.io:8080/job/ubuntu-14.04-from-scratch/661 ran the same code:
GIT_HASH: 3426a04952eb5ee906b732b4644ce77c0d24fa7e
But they have different numbers of test_cancel_insert runs: #657 has 9 and #661 has 5, leading to #657 reporting:
=============== 1624 tests deselected by "-m 'execute_serially'" =============== 187 passed, 8 skipped, 1624 deselected, 1 xfailed, 24 warnings in 1944.18 seconds
While #661 reports:
=============== 1624 tests deselected by "-m 'execute_serially'" =============== 183 passed, 8 skipped, 1624 deselected, 1 xfailed, 24 warnings in 1958.31 seconds