[ 
https://issues.apache.org/jira/browse/YARN-4754?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15177506#comment-15177506
 ] 

Rohith Sharma K S commented on YARN-4754:
-----------------------------------------

bq. I still see 2 places where we are not closing ClientResponse, when we call 
putDomain and in doPosting if response is not 200 OK.
It looks to be this is the case. After RM recovery completes, timeline entities 
are published in background. During this span of time, if there timeline sever 
is restarted or down for sometime, it is able to see many connections are kept 
CLOSE_WAIT state.

> Too many connection opened to TimelineServer while publishing entities
> ----------------------------------------------------------------------
>
>                 Key: YARN-4754
>                 URL: https://issues.apache.org/jira/browse/YARN-4754
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Rohith Sharma K S
>            Priority: Critical
>         Attachments: ConnectionLeak.rar
>
>
> It is observed that there are too many connections are kept opened to 
> TimelineServer while publishing entities via SystemMetricsPublisher. This 
> cause sometimes resource shortage for other process or RM itself
> {noformat}
> tcp        0      0 10.18.99.110:3999       10.18.214.60:59265      
> ESTABLISHED 115302/java         
> tcp        0      0 10.18.99.110:25001      :::*                    LISTEN    
>   115302/java         
> tcp        0      0 10.18.99.110:25002      :::*                    LISTEN    
>   115302/java         
> tcp        0      0 10.18.99.110:25003      :::*                    LISTEN    
>   115302/java         
> tcp        0      0 10.18.99.110:25004      :::*                    LISTEN    
>   115302/java         
> tcp        0      0 10.18.99.110:25005      :::*                    LISTEN    
>   115302/java         
> tcp        1      0 10.18.99.110:48866      10.18.99.110:8188       
> CLOSE_WAIT  115302/java         
> tcp        1      0 10.18.99.110:48137      10.18.99.110:8188       
> CLOSE_WAIT  115302/java         
> tcp        1      0 10.18.99.110:47553      10.18.99.110:8188       
> CLOSE_WAIT  115302/java         
> tcp        1      0 10.18.99.110:48424      10.18.99.110:8188       
> CLOSE_WAIT  115302/java         
> tcp        1      0 10.18.99.110:48139      10.18.99.110:8188       
> CLOSE_WAIT  115302/java         
> tcp        1      0 10.18.99.110:48096      10.18.99.110:8188       
> CLOSE_WAIT  115302/java         
> tcp        1      0 10.18.99.110:47558      10.18.99.110:8188       
> CLOSE_WAIT  115302/java         
> tcp        1      0 10.18.99.110:49270      10.18.99.110:8188       
> CLOSE_WAIT  115302/java         
> {noformat}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to