spark-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (SPARK-1202) Add a "cancel" button in the UI for stages
Date Sat, 29 Mar 2014 22:50:15 GMT

    [ https://issues.apache.org/jira/browse/SPARK-1202?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13954498#comment-13954498
] 

ASF GitHub Bot commented on SPARK-1202:
---------------------------------------

Github user kayousterhout commented on a diff in the pull request:

    https://github.com/apache/spark/pull/246#discussion_r11095610
  
    --- Diff: core/src/main/scala/org/apache/spark/ui/jobs/JobProgressListener.scala ---
    @@ -116,6 +118,16 @@ private[ui] class JobProgressListener(conf: SparkConf) extends SparkListener
{
     
         val stages = poolToActiveStages.getOrElseUpdate(poolName, new HashMap[Int, StageInfo]())
         stages(stage.stageId) = stage
    +    
    +    // Extract Job ID and double check if we have the details
    +    val jobId = Option(stageSubmitted.properties).flatMap {
    +      p => Option(p.getProperty("spark.job.id"))
    +    }.getOrElse("-1").toInt
    --- End diff --
    
    When we chatted about this I remember you saying that this code is to handle the case
where the stage runs locally at the driver...but from glancing at the DAGScheduler code, it
looks like onStageSubmitted() never gets called for the locally-run tasks, and they never
show up at the UI.
    
    When do you need this code / when will the jobIdToStageIds mapping not already be set
up correctly by OnJobStart?


> Add a "cancel" button in the UI for stages
> ------------------------------------------
>
>                 Key: SPARK-1202
>                 URL: https://issues.apache.org/jira/browse/SPARK-1202
>             Project: Apache Spark
>          Issue Type: New Feature
>          Components: Web UI
>            Reporter: Patrick Cogan
>            Assignee: Sundeep Narravula
>            Priority: Critical
>             Fix For: 1.0.0
>
>
> Seems like this would be really useful for people. It's not that hard, we just need to
lookup the jobs associated with the stage and kill them. Might involve exposing some additional
API's in SparkContext.



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

Mime
View raw message