[
https://issues.apache.org/jira/browse/HDDS-10311?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17825939#comment-17825939
]
Attila Doroszlai commented on HDDS-10311:
-----------------------------------------
bq. any idea why getMetrics in two tests will be NullPointer
This is caused by {{cluster.restartOzoneManager()}} call near the end of test
cases. Stopping Ozone Manager unregisters {{OMMetrics}}, but
{{OzoneManager.restart()}} does not re-register it.
This part of the test verifies that number of volumes/buckets/etc. is
initialized upon startup from DB. But {{OzoneManager.restart()}} is only used
by tests, and it duplicates some code from {{start()}}. So the test would not
catch any problem with missing metrics init in {{start()}}, since it exercises
another code path.
The test also injects some failures:
{code}
// inject exception to test for Failure Metrics on the read path
doThrow(exception).when(mockVm).getVolumeInfo(any());
doThrow(exception).when(mockVm).listVolumes(any(), any(),
any(), anyInt());
HddsWhiteboxTestUtils.setInternalState(ozoneManager,
"volumeManager", mockVm);
// inject exception to test for Failure Metrics on the write path
mockWritePathExceptions(OmVolumeArgs.class);
{code}
Without restarting Ozone Manager, these kinds of failures would cause problems
for the other test cases. We could fix that by restoring the original internal
state. However, I think injecting such failures is overkill. For the read
path, we can simply perform ops without creating the object in the first place.
Another problem with {{TestOmMetrics}} is that some of the expected metrics
values are "magic".
If we can add assertions about each OM operation in more specific tests
(preferably unit tests), then {{TestOmMetrics}} can be deleted and we don't
need to work around the issues above (NPE, restoring state, etc.)
> Speed up TestOmMetrics
> ----------------------
>
> Key: HDDS-10311
> URL: https://issues.apache.org/jira/browse/HDDS-10311
> Project: Apache Ozone
> Issue Type: Sub-task
> Components: test
> Reporter: Attila Doroszlai
> Assignee: Will Xiao
> Priority: Major
>
> {code}
> Tests run: 8, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 115.365 s -
> in org.apache.hadoop.ozone.om.TestOmMetrics
> {code}
> # assertions about metrics should be merged into functional tests (no need to
> test e.g. {{listKeys}} and metrics for {{listKeys}} separately
> # no need to start new cluster for each test case
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
