[ 
https://issues.apache.org/jira/browse/FLINK-30238?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17689588#comment-17689588
 ] 

Yun Gao commented on FLINK-30238:
---------------------------------

Hi [~pnowojski] , for this specific issue, based on my understanding, 
 # It looks to me the first issue might refer to the situation that on 
endOfInput(), the Writer will emit the summary message tag with \{checkpoint id 
= MAX_VALUE} to the Committer, followed by the remaining Committables. But with 
the current process, endOfInput() should not be called on stop-with-savepoint 
without draining, thus sorry I'm also not fully understand this part. 
 # For the second issue, it looks to me it refers to how the last piece of the 
committables should be passed to the post-committer topology, as posted in the 
above. Although here it refers to stop-with-savepoint, but it looks to me it 
should be stop-with-savepoint –-drain, since for stop-with-savepoint, the last 
piece of Committables should be still recorded in the last savepoint, and they 
will be re-committed after restarted. 

Perhaps [~fpaul] could give some more details?

> Unified Sink committer does not clean up state on final savepoint
> -----------------------------------------------------------------
>
>                 Key: FLINK-30238
>                 URL: https://issues.apache.org/jira/browse/FLINK-30238
>             Project: Flink
>          Issue Type: Bug
>          Components: Connectors / Common
>    Affects Versions: 1.17.0, 1.15.3, 1.16.1
>            Reporter: Fabian Paul
>            Priority: Critical
>
> During stop-with-savepoint the committer only commits the pending 
> committables on notifyCheckpointComplete.
> This has several downsides.
>  * Last committableSummary has checkpoint id LONG.MAX and is never cleared 
> from the state leading to that stop-with-savepoint does not work when the 
> pipeline recovers from a savepoint 
>  * While the committables are committed during stop-with-savepoint they are 
> not forwarded to post-commit topology, potentially losing data and preventing 
> to close open transactions.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to