[
https://issues.apache.org/jira/browse/FLINK-29637?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17645979#comment-17645979
]
David Morávek commented on FLINK-29637:
---------------------------------------
My initial impression is that this could introduce some
user-classloader-related memory leaks because the task threads can reference
these.
Do you have any thoughts on whether this could be the case?
> Improve to reuse threads in TaskManager for different tasks between jobs
> ------------------------------------------------------------------------
>
> Key: FLINK-29637
> URL: https://issues.apache.org/jira/browse/FLINK-29637
> Project: Flink
> Issue Type: Sub-task
> Components: Runtime / Task
> Affects Versions: 1.17.0
> Reporter: Shammon
> Priority: Major
>
> Currently `TaskManager` will create multiple threads for each task, such as
> execution thread, cancel thread, source thread, flush thread and etc. These
> threads will be destroyed when the task is finished. In olap cluster, the
> `TaskManager` frequently creates and destroys threads, which causes a great
> waste of resources. We should improve the `TaskManager` to reuse these thread
> after the task is finished.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)