Zita Dombi created HDDS-11063:
---------------------------------
Summary: TestSnapshotDiffManager#testThreadPoolIsFull is flaky
when there is no wait between the batches
Key: HDDS-11063
URL: https://issues.apache.org/jira/browse/HDDS-11063
Project: Apache Ozone
Issue Type: Bug
Reporter: Zita Dombi
While working on HDDS-10604 I did changes in the OzoneConfiguration
initialisation. With my changes the time it takes increased (compliance check
related changes), which made the TestSnapshotDiffManager.testThreadPoolIsFull
fail with the scenario where we don’t wait between the batches.
This shed light on the flakiness of this test. The problem is that in the
testThreadPoolIsFull method, when we there is no wait between the job batches.
Initially we create 10 snapshots. After that we create 45 snapshotdiffs and
when we are submitting the jobs (first 9, after that 8, 7, …) we are calling
the SnapshotDiffManage.getSnapshotDiffReport method. In that method we call
getSnapshotRootPath, which create an OFSPath with a new OzoneConfiguration
object each time. This will take more time than before, which makes the
assertion fail, as the 45 jobs smaller proportion will be in progress.
It’d be good to check if we can manipulate this without the wait and somehow
deterministically create a similar environment. We could also check if we can
remove that OzoneConfiguration initialisation from the OFSPath constructor.
Previously it was asserted that 20 jobs will be in progress and 25 will be
rejected, it’d be good to know how did we know to assert that.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
