[
https://issues.apache.org/jira/browse/SPARK-6641?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14566741#comment-14566741
]
Josh Rosen commented on SPARK-6641:
-----------------------------------
Can you provide more information on the motivation for this configuration?
Lazy initialization adds some complexity so there should be a good use-case to
justify it.
> Add config or control of accumulator on python
> ----------------------------------------------
>
> Key: SPARK-6641
> URL: https://issues.apache.org/jira/browse/SPARK-6641
> Project: Spark
> Issue Type: Improvement
> Components: PySpark
> Reporter: Weizhong
> Priority: Minor
>
> Now if we init SparkContext of Python, then will create a single Accumulator
> in Java and start a TCP server in a daemon thread.
> We can add config option to decide not start this TCP server at sc start,
> and start this TCP server when we use accumulator.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
