[
https://issues.apache.org/jira/browse/HUDI-7132?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17791873#comment-17791873
]
Bo Cui commented on HUDI-7132:
------------------------------
and another scenario:：
t1: chk is complete and all Coordinators start executing.
t2: flink thread pool executes the
StreamWriteOperatorCoordinator#notifyCheckpointComplete.
t3: Hudi Coordinator receives the failure message before the
StreamWriteOperatorCoordinator executes to the L524 of `commitInstant`, and
updates the value of eventBuffer[i] to null.
so eventBuffer[i] is missing from `writeResults` (L524)
> Data may be lost in flink#chk
> -----------------------------
>
> Key: HUDI-7132
> URL: https://issues.apache.org/jira/browse/HUDI-7132
> Project: Apache Hudi
> Issue Type: Bug
> Components: flink
> Affects Versions: 1.1.0
> Reporter: Bo Cui
> Priority: Major
>
> https://github.com/apache/hudi/blob/a1afcdd989ce2d634290d1bd9e099a17057e6b4d/hudi-flink-datasource/hudi-flink/src/main/java/org/apache/hudi/sink/StreamWriteOperatorCoordinator.java#L524C23-L524C35
> before the line code, eventBuffer may be updated by `subtaskFailed`, and some
> elements of eventBuffer is null
> https://github.com/apache/hudi/blob/a1afcdd989ce2d634290d1bd9e099a17057e6b4d/hudi-flink-datasource/hudi-flink/src/main/java/org/apache/hudi/sink/StreamWriteOperatorCoordinator.java#L305C10-L305C21
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
