[ 
https://issues.apache.org/jira/browse/SPARK-25210?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Hyukjin Kwon resolved SPARK-25210.
----------------------------------
    Resolution: Invalid

Let's reopen when it's clear if that's an issue.

> spark driver apply task success info cost much time 
> ----------------------------------------------------
>
>                 Key: SPARK-25210
>                 URL: https://issues.apache.org/jira/browse/SPARK-25210
>             Project: Spark
>          Issue Type: Bug
>          Components: Scheduler
>    Affects Versions: 2.1.0
>            Reporter: wangminfeng
>            Priority: Major
>
> when i run spark job by standalone, i have 600 ins and every ins can run 50 
> task. 
> driver send message to executor,  executor immediately receive the message 
> and run task.
> but when executor finish the task, send messge to driver,  driver waits much 
> time to get the messge, like this:
> driver start task,
> 18/08/23 20:20:11 DEBUG 
> scheduler.cluster.YarnSchedulerBackend$YarnDriverEndpoint 
> [dispatcher-event-loop-37]: Launching task 345733 on executor id: 261 
> hostname: yq01-inf-wangzeyu081.yq01.baidu.com
> executor start task,
> 18/08/23 20:20:11 INFO CoarseGrainedExecutorBackend: Got assigned task 345733
> there is no delay
>  
> executor finish task,
> 18/08/23 20:20:17 INFO Executor: Finished task 79.0 in stage 96.0 (TID 
> 345733). 1573 bytes result sent to driver
> 18/08/23 20:20:23 INFO spark.scheduler.TaskSetManager 
> [task-result-getter-19]: Finished task 79.0 in stage 96.0 (TID 345733) in 
> 15243 ms on yq01-inf-wangzeyu081.yq01.baidu.com (executor 261) (33095/62500)
> there is 6s delay
>  
> it is sure that no net problem.
>  
> so, how can reduce this delay



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

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

Reply via email to