[
https://issues.apache.org/jira/browse/CASSANDRA-10098?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14702120#comment-14702120
]
Paulo Motta commented on CASSANDRA-10098:
-----------------------------------------
This
[dtest|https://github.com/riptano/cassandra-dtest/blob/master/commitlog_test.py#L220]
has been flakey for quite some time in
[3.0|http://cassci.datastax.com/view/cassandra-3.0/job/cassandra-3.0_dtest/lastCompletedBuild/testReport/junit/commitlog_test/TestCommitLog/small_segment_size_test/history],
[2.2|http://cassci.datastax.com/view/cassandra-2.2/job/cassandra-2.2_dtest/lastCompletedBuild/testReport/junit/commitlog_test/TestCommitLog/small_segment_size_test/history/],
[2.1|http://cassci.datastax.com/view/cassandra-2.1/job/cassandra-2.1_dtest/lastCompletedBuild/testReport/junit/commitlog_test/TestCommitLog/small_segment_size_test/history/]
and [3.0
Windows|http://cassci.datastax.com/view/cassandra-3.0/job/cassandra-3.0_dtest_win32/17/testReport/junit/commitlog_test/TestCommitLog/small_segment_size_test/history/].
In 3.0 this been failing because it expects 13 commit log files to be generated
with an error of 20% (11 or more files), but it sometimes creates only 10
files, causing the test to fail. I wasn't able to track any particular commit
that triggered this behavior, so I'm not sure if this is a normal variation or
a test problem.
Increasing the commit log file count error tolerance from 20% to 25% makes this
test pass, but I'm not sure this is not hiding something else. It would be nice
to validate with the test author or a commit log expert if this variation is
acceptable. My guess is, that with faster hardware and other optimizations,
it's possible to write less commit log files for the same amount of data, so
it's safe to increase this tolerance.
The cassandra-dtest PR with the fix is available on this
[link|https://github.com/riptano/cassandra-dtest/pull/477]
> Windows dtest 3.0: commitlog_test.py:TestCommitLog.small_segment_size_test
> fails
> --------------------------------------------------------------------------------
>
> Key: CASSANDRA-10098
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10098
> Project: Cassandra
> Issue Type: Sub-task
> Reporter: Joshua McKenzie
> Assignee: Paulo Motta
> Labels: Windows
> Fix For: 3.0.x
>
>
> {noformat}
> File "D:\Python27\lib\unittest\case.py", line 329, in run
> testMethod()
> File
> "D:\jenkins\workspace\cassandra-3.0_dtest_win32\cassandra-dtest\tools.py",
> line 243, in wrapped
> f(obj)
> File
> "D:\jenkins\workspace\cassandra-3.0_dtest_win32\cassandra-dtest\commitlog_test.py",
> line 226, in small_segment_size_test
> self._commitlog_test(segment_size_in_mb, 62.5, 13, files_error=0.2)
> File
> "D:\jenkins\workspace\cassandra-3.0_dtest_win32\cassandra-dtest\commitlog_test.py",
> line 99, in _commitlog_test
> error=files_error)
> File
> "D:\jenkins\workspace\cassandra-3.0_dtest_win32\cassandra-dtest\assertions.py",
> line 62, in assert_almost_equal
> assert vmin > vmax * (1.0 - error) or vmin == vmax, "values not within
> %.2f%% of the max: %s" % (error * 100, args)
> 'values not within 20.00% of the max: (10, 13)\n-------------------- >> begin
> captured logging << --------------------\ndtest: DEBUG: cluster ccm
> directory: d:\\temp\\dtest-qnguzs\n--------------------- >> end captured
> logging << ---------------------'
> {noformat}
> Failure history:
> [consistent|http://cassci.datastax.com/view/cassandra-3.0/job/cassandra-3.0_dtest_win32/17/testReport/junit/commitlog_test/TestCommitLog/small_segment_size_test/]
> env: Both ci and local
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
