[
https://issues.apache.org/jira/browse/FLINK-25850?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roman Khachatryan updated FLINK-25850:
--------------------------------------
Priority: Minor (was: Major)
> Consider notifying nested state backend about checkpoint abortion
> -----------------------------------------------------------------
>
> Key: FLINK-25850
> URL: https://issues.apache.org/jira/browse/FLINK-25850
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / State Backends
> Reporter: Roman Khachatryan
> Priority: Minor
> Fix For: 1.16.0
>
>
> The notification is optional, but some backends might do GC upon receiving it.
> {code}
> These notifications are "best effort", meaning they can sometimes be skipped.
> This method is very rarely necessary to implement.
> {code}
> The usefulness is also limited by:
> - low probability of notification reaching backend because of the difference
> in intervals and cleanup on checkpoint completion
> - low probability of backends making good use of it because it's delivered
> after snapshot is done; and backends must be resilient to missing
> notifications
> There is added complexity and risk (such as FLINK-25816).
> Probably, complexity can be eliminated by extracting some Notifier class from
> ChangelogStateBackend.
> cc: [~yunta]
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
