[
https://issues.apache.org/jira/browse/SPARK-27192?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sean Owen resolved SPARK-27192.
-------------------------------
Resolution: Fixed
Issue resolved by pull request 24261
[https://github.com/apache/spark/pull/24261]
> spark.task.cpus should be less or equal than spark.task.cpus when use static
> executor allocation
> ------------------------------------------------------------------------------------------------
>
> Key: SPARK-27192
> URL: https://issues.apache.org/jira/browse/SPARK-27192
> Project: Spark
> Issue Type: Improvement
> Components: Spark Core
> Affects Versions: 2.2.0, 2.3.0, 2.4.0
> Reporter: Lijia Liu
> Assignee: Lijia Liu
> Priority: Minor
> Fix For: 3.0.0
>
>
> When use dynamic executor allocation, if we set spark.executor.cores small
> than spark.task.cpus, exception will be thrown as follows:
> '''spark.executor.cores must not be < spark.task.cpus'''
> But, if dynamic executor allocation not enabled, spark will hang when submit
> new job for TaskSchedulerImpl will not schedule a task in a executor which
> available cores is small than
> spark.task.cpus.See
> [https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala#L351]
> So, when start task scheduler, spark.task.cpus should be check.
> reproduce
> $SPARK_HOME/bin/spark-shell --conf spark.task.cpus=2 --master local[1]
> scala> sc.parallelize(1 to 9).collect
--
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
