[
https://issues.apache.org/jira/browse/CASSANDRA-16774?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17372186#comment-17372186
]
David Capwell edited comment on CASSANDRA-16774 at 6/30/21, 8:04 PM:
---------------------------------------------------------------------
looking at the test I feel more is going on as well.
* BinLog is async and max size is only used by the archiver, which is triggered
on release (I don't see how we control it)
* max weight is 10, and our records don't override weight, so 1 record = 1
weight; aka max queue size of 10 records
* test is waiting until a log exists which has 2 records... but 0, 1, 2, 3, 4,
5, 6, 7, 8, 9, and 10 are all valid states
Unless I am missing something,
org.apache.cassandra.utils.binlog.BinLogTest#flakyTestTruncationReleasesLogSpace
doesn't make sense to me. The test would make more sense if it controlled the
archiver and wrote 11 records, you can then detect release got triggered and
then the log size should be 1
But to your point, since stop doesn't release the resources the files cleanup
when GC triggered, so previous tests may have other log files in the test
directory
was (Author: dcapwell):
looking at the test I feel more is going on as well.
* BinLog is async and max size is only used by the archiver, which is triggered
on release (I don't see how we control it)
* max weight is 10, and our records don't override weight, so 1 record = 1
weight; aka max queue size of 10 records
* test is waiting until a log exists which has 2 records... but 0, 1, 2, 3, 4,
5, 6, 7, 8, 9, and 10 are all valid states
Unless I am missing something,
org.apache.cassandra.utils.binlog.BinLogTest#flakyTestTruncationReleasesLogSpace
doesn't make sense to me. The test would make more sense if it controlled the
archiver and wrote 11 records, you can then detect release got triggered and
then the log size should be 1
> Fix flaky tests auditlog_test.py::TestAuditlog caused by the fact we do not
> close chronicle
> -------------------------------------------------------------------------------------------
>
> Key: CASSANDRA-16774
> URL: https://issues.apache.org/jira/browse/CASSANDRA-16774
> Project: Cassandra
> Issue Type: Bug
> Components: Test/dtest/python, Tool/fql
> Reporter: David Capwell
> Assignee: David Capwell
> Priority: Normal
> Fix For: 4.0.1, 4.1
>
>
> auditlog_test.py::TestAuditlog::test_archiving_fql and
> test_fql_nodetool_options fail from time to time due to the test relying on a
> race condition; we do not close chronicle queue so rotation may not happen
> before stopping archiver, the tests fail if rotation happens before stopping
> archiver (which is done based off GC).
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
