[
https://issues.apache.org/jira/browse/FLINK-22108?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17314226#comment-17314226
]
Kurt Young commented on FLINK-22108:
------------------------------------
Nice catch [~kezhuw]
[~TsReaper] Any idea to improve this?
> Ephemeral socket address was checkpointed to state and restored back in
> CollectSinkOperatorCoordinator
> ------------------------------------------------------------------------------------------------------
>
> Key: FLINK-22108
> URL: https://issues.apache.org/jira/browse/FLINK-22108
> Project: Flink
> Issue Type: Bug
> Components: API / DataStream
> Affects Versions: 1.13.0
> Reporter: Kezhu Wang
> Priority: Major
>
> {{CollectSinkOperatorCoordinator}} checkpointed its {{address}} field to
> state and restored back. That field is listener address of
> {{CollectSinkFunction}}. After {{resetToCheckpoint}} (eg. global failover}},
> {{address}} is meaningless. If client request comes before
> {{CollectSinkAddressEvent}}, it will use that meaningless address for
> connection. In best situation, error happens, and nothing hurts. In bad
> situation, no one knows where the restored address points to now.
> cc [~TsReaper] [~ykt836]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
