[jira] [Commented] (FLINK-5857) Recycle idle containers in time for yarn mode

["Till Rohrmann (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-5857?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15882800#comment-15882800 ] 

Till Rohrmann commented on FLINK-5857:
--------------------------------------

Hi [~tiemsn], the new {{SlotManager}} implementation (see FLINK-5810) introduces a timeout for idling {{TaskExecutors}}. Once a {{TaskExecutor}} times out, the {{TaskExecutor}} is removed from the {{SlotManager}} and the {{ResourceManager}} is informed about it (e.g. to release the container). 

Thus, I think we can close this issue. What do you think?

> Recycle idle containers in time for yarn mode
> ---------------------------------------------
>
>                 Key: FLINK-5857
>                 URL: https://issues.apache.org/jira/browse/FLINK-5857
>             Project: Flink
>          Issue Type: Bug
>          Components: YARN
>            Reporter: shuai.xu
>            Assignee: shuai.xu
>              Labels: flip-6
>
> When we run flink batch job like map reduce, after a map is finished, the container for it may be idle for a long time, we need to have a strategy to recycle there container to reduce resource usage



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)