[
https://issues.apache.org/jira/browse/FLINK-14979?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16983678#comment-16983678
]
Piotr Nowojski commented on FLINK-14979:
----------------------------------------
One extra question from my side. If following checkpoints do not subsume the
savepoints, should such checkpoint completed notification subsume the savepoint
as well? What if the savepoint fails? I think, at least in this case, this is a
non-issue, since side effects for both the savepoint and checkpoint will be
committed by the checkpoint's complete notification.
> TwoPhaseCommitSink fails when checkpoint overtakes a savepoint
> --------------------------------------------------------------
>
> Key: FLINK-14979
> URL: https://issues.apache.org/jira/browse/FLINK-14979
> Project: Flink
> Issue Type: Bug
> Components: API / DataStream
> Affects Versions: 1.6.4, 1.7.2, 1.8.2, 1.9.1
> Reporter: Piotr Nowojski
> Priority: Major
>
> As [reported by a user on the user mailing
> list|http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/UNCHECKED-Error-while-confirming-Checkpoint-td23213.html],
> {{TwoPhaseCommitSinkFunction#notifyCheckpointComplete}} can fail with the
> following exception:
> {noformat}
> java.lang.RuntimeException: Error while confirming checkpoint
> at org.apache.flink.runtime.taskmanager.Task$2.run(Task.java:1205)
> at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
> at java.util.concurrent.FutureTask.run(FutureTask.java:266)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> at java.lang.Thread.run(Thread.java:748)
> Caused by: java.lang.IllegalStateException: checkpoint completed, but no
> transaction pending
> at org.apache.flink.util.Preconditions.checkState(Preconditions.java:195)
> at
> org.apache.flink.streaming.api.functions.sink.TwoPhaseCommitSinkFunction.notifyCheckpointComplete(TwoPhaseCommitSinkFunction.java:267)
> at
> org.apache.flink.streaming.api.operators.AbstractUdfStreamOperator.notifyCheckpointComplete(AbstractUdfStreamOperator.java:130)
> at
> org.apache.flink.streaming.runtime.tasks.StreamTask.notifyCheckpointComplete(StreamTask.java:822)
> at org.apache.flink.runtime.taskmanager.Task$2.run(Task.java:1200)
> ... 5 more
> {noformat}
> 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
> In this case, {{TwoPhaseCommitSinkFunction}} receives first notification that
> the later checkpoint completed, it commits both savepoint and the checkpoint.
> Later when savepoint {{notifyCheckpointComplete}} arrives, the above error
> will occur.
> Possible trivial fix is to remove that failing {{checkState}}.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
