[
https://issues.apache.org/jira/browse/FLINK-17978?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17130586#comment-17130586
]
Till Rohrmann commented on FLINK-17978:
---------------------------------------
Hi [~fly_in_gis], thanks for helping with this testing task.
The idea of the testing task is that Flink can properly run with a Hadoop
dependency which is not {{flink-shaded-hadoop-*.jar}}. This basically tests
that we don't have any class loading issues by incorrectly bundled
dependencies. In order to test it, it would be a good idea to select a couple
of Hadoop versions (e.g. what you have proposed) and run a Flink job which
directly/indirectly uses it. One idea could be to run a job writing to HDFS.
Another idea was to use HBase or ORC for the interaction with Hadoop.
Maybe we could start with simply writing and reading from HDFS for different
Hadoop versions.
> Test Hadoop dependency change
> -----------------------------
>
> Key: FLINK-17978
> URL: https://issues.apache.org/jira/browse/FLINK-17978
> Project: Flink
> Issue Type: Sub-task
> Components: Runtime / Coordination
> Affects Versions: 1.11.0
> Reporter: Till Rohrmann
> Assignee: Yang Wang
> Priority: Critical
> Labels: release-testing
> Fix For: 1.11.0
>
>
> Test the Hadoop dependency change:
> * Run Flink with HBase/ORC (maybe add e2e test)
> * Validate meaningful exception message if Hadoop dependency is missing
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
