[
https://issues.apache.org/jira/browse/FLINK-4960?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15633188#comment-15633188
]
ASF GitHub Bot commented on FLINK-4960:
---------------------------------------
Github user aljoscha commented on a diff in the pull request:
https://github.com/apache/flink/pull/2747#discussion_r86372111
--- Diff:
flink-streaming-java/src/test/java/org/apache/flink/streaming/util/AbstractStreamOperatorTestHarness.java
---
@@ -309,6 +310,85 @@ public void initializeState(OperatorStateHandles
operatorStateHandles) throws Ex
initializeCalled = true;
}
+ /**
+ * Takes the different {@link OperatorStateHandles} created by calling
{@link #snapshot(long, long)}
+ * on different instances of {@link AbstractStreamOperatorTestHarness}
(each one representing one subtask)
+ * and repacks them into a single {@link OperatorStateHandles} so that
the parallelism of the test
+ * can change arbitrarily (i.e. be able to scale both up and down).
+ * <p/>
--- End diff --
A new paragraph should start with `<p>`. `<p/>` should never be used in
Javadoc, AFAIK.
> Allow the AbstractStreamOperatorTestHarness to test scaling down
> ----------------------------------------------------------------
>
> Key: FLINK-4960
> URL: https://issues.apache.org/jira/browse/FLINK-4960
> Project: Flink
> Issue Type: New Feature
> Components: Tests
> Reporter: Kostas Kloudas
> Assignee: Kostas Kloudas
> Fix For: 1.2.0
>
>
> Currently the AbstractStreamOperatorTestHarness allows for testing an
> operator when scaling up, through snapshot and restore. This is not enough
> as many interesting corner cases arise when scaling down or during
> arbitrary combinations of scaling up and down.
> This issue targets to add this functionality so that an operator can snapshot
> its state, restore with different parallelism, and later scale down or
> further up.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
