You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Andrew Ash (JIRA)" <ji...@apache.org> on 2014/11/14 10:51:33 UTC

[jira] [Commented] (SPARK-809) Give newly registered apps a set of executors right away

    [ https://issues.apache.org/jira/browse/SPARK-809?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14212068#comment-14212068 ] 

Andrew Ash commented on SPARK-809:
----------------------------------

I believe this situation hasn't changed?

Looking a little at the master branch (1.2) the standalone cluster still gets its {{defaultParallelism}} in {{CoarseGrainedSchedulerBackend}} from the {{totalCoreCount}} that is incremented/decremented as executors join/leave the driver's awareness.

> Give newly registered apps a set of executors right away
> --------------------------------------------------------
>
>                 Key: SPARK-809
>                 URL: https://issues.apache.org/jira/browse/SPARK-809
>             Project: Spark
>          Issue Type: New Feature
>          Components: Deploy
>            Reporter: Matei Zaharia
>            Priority: Minor
>
> Right now, newly connected apps in the standalone cluster will not set a good defaultParallelism value if they create RDDs right after creating a SparkContext, because the executorAdded calls are asynchronous and happen after. It would be nice to wait for a few such calls before returning from the scheduler initializer.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org