-
Type:
Bug Report
-
Resolution: Fixed
-
Priority:
L3 - Default
-
Affects Version/s: None
-
Component/s: backend, continuous integration
-
None
-
Not defined
Context:
We sometimes see the Data upgrade test stage fail on branch builds due to an unexpected amount (exactly double) of event process instances during a republish test:
`com.camunda.optimize.test.upgrade.UpgradeDataIT.upgradeWithDataAndPerformRegressionTest`:
Failures (1):
JUnit Jupiter:PostMigrationTest:republishAllEventBasedProcessesAndEnsureTheyArePublishedAndHaveInstanceData()
MethodSource [className = 'com.camunda.optimize.test.upgrade.PostMigrationTest', methodName = 'republishAllEventBasedProcessesAndEnsureTheyArePublishedAndHaveInstanceData', methodParameterTypes = '']
=> org.opentest4j.AssertionFailedError:
expected: {"46ade747-84f2-4b40-85a2-fbdbef73543e"=3L, "7b917036-cd64-4e39-9464-7d4457adf51b"=10L}
but was : {"46ade747-84f2-4b40-85a2-fbdbef73543e"=6L, "7b917036-cd64-4e39-9464-7d4457adf51b"=10L}
at com.camunda.optimize.test.upgrade.UpgradeDataIT.upgradeWithDataAndPerformRegressionTest(UpgradeDataIT.groovy:53)
AT:
- identify the cause and either adjust the test or the publish logic to avoid the potential race-condition
Hint:
This may be caused by a race condition between the actual process republish call as the new publish state is created before the old one is deleted and the `EventBasedProcessesInstanceImportScheduler` may run just in between that time window when both exist at the same time.