[
https://issues.apache.org/jira/browse/FLINK-32810?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Zhipeng Zhang resolved FLINK-32810.
-----------------------------------
Fix Version/s: ml-2.4.0
Assignee: Fan Hong
Resolution: Fixed
> Improve managed memory usage in ListStateWithCache
> --------------------------------------------------
>
> Key: FLINK-32810
> URL: https://issues.apache.org/jira/browse/FLINK-32810
> Project: Flink
> Issue Type: Improvement
> Components: Library / Machine Learning
> Reporter: Fan Hong
> Assignee: Fan Hong
> Priority: Major
> Labels: pull-request-available
> Fix For: ml-2.4.0
>
>
> Right now, by default, an instance of `ListStateWithCache` uses up all the
> managed memory allocated for`ManagedMemoryUseCase.OPERATOR`.
> It could bring some bugs in some situations, for example, when there exist
> more than one `ListStateWithCache`s in a single operator, or there are other
> places using managed memory of `ManagedMemoryUseCase.OPERATOR`.
>
> An approach to resolve such cases is to let the developer be aware about the
> usage of managed memory of `ManagedMemoryUseCase.OPERATOR`, instead of
> implicitly use up all of it. Therefore, I think it is better to add a
> parameters `fraction` to specify how much memory is used in the
> `ListStateWithCache`.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
