[
https://issues.apache.org/jira/browse/FLINK-2491?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16936570#comment-16936570
]
Tomas Witzany commented on FLINK-2491:
--------------------------------------
Hi, came across this issue. It is especially unintuitive whenever working with
broadcast state. In that case the typical use case would be load the broadcast
stream from somewhere in a source and broadcast it. Why is it necessary to keep
it alive?
> Checkpointing only works if all operators/tasks are still running
> -----------------------------------------------------------------
>
> Key: FLINK-2491
> URL: https://issues.apache.org/jira/browse/FLINK-2491
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Checkpointing
> Affects Versions: 0.10.0
> Reporter: Robert Metzger
> Priority: Critical
> Attachments: fix_checkpoint_not_working_if_tasks_are_finished.patch
>
>
> While implementing a test case for the Kafka Consumer, I came across the
> following bug:
> Consider the following topology, with the operator parallelism in parentheses:
> Source (2) --> Sink (1).
> In this setup, the {{snapshotState()}} method is called on the source, but
> not on the Sink.
> The sink receives the generated data.
> only one of the two sources is generating data.
> I've implemented a test case for this, you can find it here:
> https://github.com/rmetzger/flink/blob/para_checkpoint_bug/flink-tests/src/test/java/org/apache/flink/test/checkpointing/ParallelismChangeCheckpoinedITCase.java
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
