[
https://issues.apache.org/jira/browse/FLINK-29509?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17616820#comment-17616820
]
Fabian Paul commented on FLINK-29509:
-------------------------------------
Merge into
master: 7a509c46e45b9a91f2b7d01f13afcdef266b1faf
release-1.16: d51389dc33af21038c982b733b86af8bbb736d19
release-1.15: 6b4882791dd0fd1b0df952ed7712ae7bd68adf36
> Set correct subtaskId during recovery of committables
> -----------------------------------------------------
>
> Key: FLINK-29509
> URL: https://issues.apache.org/jira/browse/FLINK-29509
> Project: Flink
> Issue Type: Bug
> Components: Connectors / Common
> Affects Versions: 1.16.0, 1.17.0, 1.15.2
> Reporter: Fabian Paul
> Assignee: Krzysztof Chmielewski
> Priority: Critical
> Labels: pull-request-available
>
> When we recover the `CheckpointCommittableManager` we ignore the subtaskId it
> is recovered on.
> [https://github.com/apache/flink/blob/d191bda7e63a2c12416cba56090e5cd75426079b/flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/operators/sink/committables/CheckpointCommittableManagerImpl.java#L58]
> This becomes a problem when a sink uses a post-commit topology because
> multiple committer operators might forward committable summaries coming from
> the same subtaskId.
>
> It should be possible to use the subtaskId already present in the
> `CommittableCollector` when creating the `CheckpointCommittableManager`s.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
