[
https://issues.apache.org/jira/browse/FLINK-19377?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17200820#comment-17200820
]
Arvid Heise commented on FLINK-19377:
-------------------------------------
I dug into it and see it as a non-issue. Here is what happens:
* You have a failing source that fails once.
* JM cancels job.
* During cleanup, sink verifies and fails as well.
* That failure is hidden because task is failing already.
* Restart happens (restarts = 1 in strategy).
* Job finishes normally.
* During cleanup, sink verifies but does not fail in heap backends.
* Test succeeds.
Not sure why your test succeeds in the second attempt, but that should be easy
to debug.
> Task can swallow test exceptions which hides test failures
> ----------------------------------------------------------
>
> Key: FLINK-19377
> URL: https://issues.apache.org/jira/browse/FLINK-19377
> Project: Flink
> Issue Type: Bug
> Components: Runtime / State Backends, Runtime / Task, Tests
> Affects Versions: 1.8.3, 1.9.3, 1.10.2, 1.11.2
> Reporter: Aljoscha Krettek
> Priority: Blocker
> Fix For: 1.12.0, 1.11.3
>
>
> I noticed this while debugging {{EventTimeWindowCheckpointingITCase}}. If you
> change {{testSlidingWindows()}} to use a tumbling window instead of the
> sliding window the test should fail. Interestingly, the tests only fail when
> using the RocksDB backend but not for the heap-based backends.
> You can follow the flow by setting a breakpoint on the {{AssertionError}}
> being thrown in {{ValidatingSink}}. Somewhere in the Task innards it will be
> suppressed/swallowed and the test will succeed.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
