[
https://issues.apache.org/jira/browse/NIFI-6184?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16811401#comment-16811401
]
Matt Burgess commented on NIFI-6184:
------------------------------------
If an SSL error occurs, I would expect a checked exception (or perhaps an
unchecked one) that should be caught and handled correctly, meaning when the
run() method is called on the processor-under-test, it should propagate up an
exception or route the flow file to failure. I will add an assertion that there
are no failed flow files, knowing that if for some reason an SSL error occurs
(and the flow file is transferred to failure), the test (and build) will fail
rather than run indefinitely.
However I think there's another issue here with the Travis build running
indefinitely, and that is a race condition between the publishing of the
nifi-standard-processors test-jar artifact, and the resolution by the
nifi-livy-processors module. These events seem (at least lately) to be
occurring in parallel, causing most/all Travis CI builds to timeout. I looked
into this and noticed that nifi-livy-processors does not depend on anything in
the nifi-standard-processors test-jar, so we should just remove that dependency
(and associated possible race condition). Note that the same race condition
could happen with any other parallel goal execution, if the other module's goal
includes resolving the nifi-standard-processors test-jar dependency. We may
want to update the POM and/or .travis.yml to ensure that artifact is published
before other NAR modules are built.
> ExecuteSparkInteractiveTestBase tests can run indefinitely
> ----------------------------------------------------------
>
> Key: NIFI-6184
> URL: https://issues.apache.org/jira/browse/NIFI-6184
> Project: Apache NiFi
> Issue Type: Bug
> Components: Extensions
> Affects Versions: 1.9.1
> Reporter: Jeff Storck
> Assignee: Matt Burgess
> Priority: Major
>
> Tests extending ExecuteSparkInteractiveTestBase can enter a state in which
> they never complete, effectively keeping the test from completing if flowfile
> processing has failed in a way that ends with the flowfile being returned to
> the incoming queue.
> For instance, if there is an issue preventing an SSL connection from being
> made, the flowfile will not be taken from the incoming queue, preventing the
> while loop in ExecuteSparkInteractiveTestBase from reaching its terminating
> condition.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
