flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Fabian Hueske (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (FLINK-1017) Add setParallelism() to Java API documentation
Date Fri, 15 Aug 2014 07:57:18 GMT

    [ https://issues.apache.org/jira/browse/FLINK-1017?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14098318#comment-14098318
] 

Fabian Hueske commented on FLINK-1017:
--------------------------------------

With {{env.setDefaultLocalParallelism(3)} each task of a job will be executed with a DOP of
3 if no parallelism is specified on task-level.

You are right, that this will result in three parallel task pipelines of DataSource ->
Map -> Reduce -> DataSink.
If you increase the DOP of the DataSink ({{counts.print().setParallelism(4);}}, the result
will be written with four threads (0>,1>,2>, and 3>).

It is not possible to specify which keys should be processed by which thread. 

> Add setParallelism() to Java API documentation
> ----------------------------------------------
>
>                 Key: FLINK-1017
>                 URL: https://issues.apache.org/jira/browse/FLINK-1017
>             Project: Flink
>          Issue Type: Task
>          Components: Documentation
>    Affects Versions: 0.6-incubating, pre-apache-0.5
>            Reporter: Fabian Hueske
>            Assignee: Hung Chang
>            Priority: Minor
>              Labels: starter
>             Fix For: 0.6-incubating
>
>
> The Java API offers {{setParallelism()}} to control the degree of parallelism for each
operator. This feature is not documented and should be added.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message