Huang Kuan Hao created HDDS-16223:
-------------------------------------
Summary: TestRefreshVolumeUsageHandler#test is flaky
Key: HDDS-16223
URL: https://issues.apache.org/jira/browse/HDDS-16223
Project: Apache Ozone
Issue Type: Sub-task
Reporter: Huang Kuan Hao
Failed here (integration (container) job of a fork CI run for an unrelated
change), then passed on rerun of the same job, so this is flaky:
https://github.com/rich7420/ozone/actions/runs/32114853096/job/95646137582
[ERROR] Tests run: 1, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 26.11 s
<<< FAILURE! -- in
org.apache.hadoop.ozone.container.common.statemachine.commandhandler.TestRefreshVolumeUsageHandler
[ERROR] TestRefreshVolumeUsageHandler.test:105 expected: <true> but was: <false>
org.opentest4j.AssertionFailedError: expected: <true> but was: <false>
at
org.apache.hadoop.ozone.container.common.statemachine.commandhandler.TestRefreshVolumeUsageHandler.test(TestRefreshVolumeUsageHandler.jav
Line 105 asserts that SCM's recorded used space still equals the pre-write
value, before the node report is expected to refresh it:
assertTrue(cluster.getStorageContainerManager().getScmNodeManager().getUsageInfo(datanodeDetails).getScmNodeStat().getScmUsed().isEqual(currentScmUsed));
This is a timing race: the test assumes the datanode's node report / DU cache
update has not yet propagated the new used space to SCM at this point, but the
async report can land before the assertion runs, making it false. The
subsequent waitFor for the refreshed value is the real check; this
pre-condition assert is the flaky one. It passes on rerun.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
