[
https://issues.apache.org/jira/browse/FLINK-36585?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17929851#comment-17929851
]
Jiangjie Qin commented on FLINK-36585:
--------------------------------------
[~Sergey Nuyanzin] Sorry for the late fix. I have identified the issue ( an
accidental change of the timeout unit from millisecond to nanoseconds). I'll
provide a patch and re-check in the fix today.
> The SplitFetcherManager.close() can chain the futures in a tight loop.
> ----------------------------------------------------------------------
>
> Key: FLINK-36585
> URL: https://issues.apache.org/jira/browse/FLINK-36585
> Project: Flink
> Issue Type: Bug
> Components: Connectors / Common
> Affects Versions: 2.0-preview
> Reporter: Jiangjie Qin
> Assignee: Jiangjie Qin
> Priority: Major
> Labels: pull-request-available
> Fix For: 2.0.0
>
>
> This is an issue introduced in FLINK-35924. The SplitFetcherManager.close()
> may chain the fetcher status check to the element queue availability future
> in a tight loop. This may cause problems such as OOM if the fetcher closure
> takes long.
> The fix is straightforward. We just need to blocking waiting on each check.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
