[
https://issues.apache.org/jira/browse/FLINK-20492?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17246933#comment-17246933
]
Jiangjie Qin commented on FLINK-20492:
--------------------------------------
Per discussion in the [PR|https://github.com/apache/flink/pull/14314]. The
correct solution here is closing the {{SourceReader}} in
{{SourceOperator.dispose()}} which will be called in all cases.
The PR has been merged:
master: 168124f99c75e873adc81437c700f85f703e2248
release-1.12: c22bb49443e6d70be5d312acee065fc9b4d95bb2
> The SourceOperatorStreamTask should implement cancelTask() and finishTask()
> ---------------------------------------------------------------------------
>
> Key: FLINK-20492
> URL: https://issues.apache.org/jira/browse/FLINK-20492
> Project: Flink
> Issue Type: Bug
> Components: Connectors / Common
> Reporter: Jiangjie Qin
> Assignee: Jiangjie Qin
> Priority: Blocker
> Labels: pull-request-available
> Fix For: 1.13.0, 1.12.1
>
>
> The {{SourceOperatorStreamTask}} does not implement {{cancelTask}} and
> {{finishTask}} at this point. This causes resource leak on job cancellation
> of finish.
> Currently there are three cases that a Job may exit.
> # Canceled - {{cancelTask()}} method will be called to close the
> SourceOperator.
> # Stopped with a savepoint - {{finishTask()}} method will be called to close
> the SourceOperator.
> # Runs to the EndOfInput - the SourceOperator is closed in
> {{StreamTask.afterInvoke()}}.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
