[
https://issues.apache.org/jira/browse/FLINK-24080?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17407308#comment-17407308
]
Chesnay Schepler commented on FLINK-24080:
------------------------------------------
Close related to FLINK-23147.
Task#doRun sets the context class loader of the executing thread to the user
code classloader, and it is still set when FileCache#releaseJob is called. This
may create threads in the thread pool that backs the FileCache, which then
continue to reference the user code classloader.
> taskmanagers leak suspect ChildFirstClassLoader Biggest instances
> -------------------------------------------------------------------
>
> Key: FLINK-24080
> URL: https://issues.apache.org/jira/browse/FLINK-24080
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Task
> Affects Versions: 1.11.0, 1.12.0, 1.13.0, 1.14.0
> Reporter: leishuiyu
> Priority: Major
> Attachments: GC root.png, class add.png, screenshot-1.png
>
>
> * Symptom After submitting a batch task for several times, the task is
> abnormal ,Caused by: java.lang.OutOfMemoryError: Direct buffer memory
> * MAT analyzing memory find five ChildFirstClassLoader instances, submit
> five batch job
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
