You are viewing a plain text version of this content. The canonical link for it is here.

Posted to builds@flink.apache.org by Flink CI <fl...@foxmail.com> on 2019/11/29 05:08:05 UTC

Broken: apache/flink#41248 (release-1.9 - cf82a14)

|  [

apache

/

flink

](https://travis-ci.org/apache/flink)  
---  
  
![branch icon](https://s3.amazonaws.com/travis-email-
assets/branch.png)[release-1.9](https://github.com/apache/flink/tree/release-1.9)  
  
|

![build hasfailed](https://s3.amazonaws.com/travis-email-assets/status-
failed.png)

[Build #41248 broken](https://travis-ci.org/apache/flink/builds/618270918)

![arrow to build time](https://s3.amazonaws.com/travis-email-assets/failure-
arrow.png)

![clock icon](https://s3.amazonaws.com/travis-email-assets/time.png)4 hrs, 33
mins, 57 secs  
  
---  
  
![Stefan Richter
avatar](https://secure.gravatar.com/avatar/17bcd3be795cc5011fbdf2c4e597ab08.jpg)Stefan
Richter

[cf82a14 CHANGESET
->](https://github.com/apache/flink/compare/e7f540c3ef39...cf82a145921d)  
  
[FLINK-10377] Support checkpoint overtaking a savepoint in TwoPhaseCommitSink  
  
The precondition checkState(pendingTransactionIterator.hasNext(), "checkpoint
completed, but no transaction pending"); in
TwoPhaseCommitSinkFunction.notifyCheckpointComplete() seems too strict,
because checkpoints can overtake checkpoints and will fail the precondition.
In this case the commit was already performed by the first notification and
subsumes the late checkpoint. I think the check can be removed.  
  
This can happen in the following scenario:  
# savepoint is triggered  
# checkpoint is triggered  
# checkpoint completes (but it doesn't subsume the savepoint, because
checkpoints subsume only other checkpoints).  
# savepoint completes