[ 
https://issues.apache.org/jira/browse/FLINK-15415?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17003922#comment-17003922
 ] 

Yun Tang commented on FLINK-15415:
----------------------------------

[~hiliuxg], generally speaking, it's not recommend to directly open a bug issue 
which is complex due to many possible causes and unclear without an explicit 
exception stack but a comprehensive unexpected behavior. The better place is to 
ask in user mail list. Especially you have asked a general problem of removed 
assigned slot in FLINK-15388.

Flink community is really active and glad to answer your question in mail list. 
If we finally figure out this is a real bug, that's the time to create a issue 
here.

> Flink machine node memory is consumed quickly, but the heap has not changed
> ---------------------------------------------------------------------------
>
>                 Key: FLINK-15415
>                 URL: https://issues.apache.org/jira/browse/FLINK-15415
>             Project: Flink
>          Issue Type: Bug
>          Components: Runtime / State Backends
>    Affects Versions: 1.8.0
>         Environment: machine : 256G memory , SSD , 32 CORE CPU
> config :
> jobmanager.heap.size: 4096m
> taskmanager.heap.size: 144gb
> taskmanager.numberOfTaskSlots: 48
> taskmanager.memory.fraction: 0.7
> taskmanager.memory.off-heap: false
> parallelism.default: 1
>  
>  
>  
>  
>  
>            Reporter: hiliuxg
>            Priority: Major
>         Attachments: flink-3.png, flink-pid-2.png, flink-pid.png
>
>
>  The Flink node has been running for 1 month, and it is found that the 
> machine's memory is getting higher and higher, even exceeding the memory 
> configured by the heap. Through top -c, it is found that the machine's memory 
> and virtual memory are very high. It may be an off-heap memory leak. What is 
> the collection mechanism of flink's off-heap memory? It need to start FULL GC 
> to recycle off-heap memory?
> In addition, does flink disable off-heap memory by default? 
> taskmanager.memory.off-heap: false, why is the memory still increasing slowly 
> .
>  



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to