[jira] [Commented] (FLINK-1518) Laggy task cancelling on task managers

["Stephan Ewen (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-1518?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15043952#comment-15043952 ] 

Stephan Ewen commented on FLINK-1518:
-------------------------------------

@uce Can you double check this issue?

I don't think this has anything to do with the web frontend. Also, not sure if this is an issue in Flink altogether. How smooth cancellation works depends on whether the cancelled code cooperates fast, or takes long to clean up after itself.

> Laggy task cancelling on task managers
> --------------------------------------
>
>                 Key: FLINK-1518
>                 URL: https://issues.apache.org/jira/browse/FLINK-1518
>             Project: Flink
>          Issue Type: Improvement
>          Components: Distributed Runtime
>    Affects Versions: 0.10.0
>            Reporter: Ufuk Celebi
>
> When cancelling a task on a cluster setup via the web interface I've noticed that task cancellation is laggy. The tasks switch to the CANCELING state, but the cancelling of the tasks takes some time.
> There are also corresponding log messages:
> {code}
> 14:29:29,281 WARN  org.apache.flink.runtime.execution.RuntimeEnvironment         - Task CHAIN DataSource (at getTextDataSet(WordCount.java:141) (org.apache.flink.api.java.io.TextInputF
> /02/11/2015 14:53:22    CHAIN DataSource (at getTextDataSet(WordCount.java:141) (org.apache.flink.api.java.io.TextInputFormat)) -> FlatMap (FlatMap at main(WordCount.java:67)) -> │ormat)) -> FlatMap (FlatMap at main(WordCount.java:67)) -> Combine(SUM(1), at main(WordCount.java:67) (7/24) did not react to cancelling signal. Sending repeated interrupt.
> {code}
> Tasks are cancelled at the task managers by repeatedly sending an interrupt signal to the task threads.



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