[
https://issues.apache.org/jira/browse/SPARK-19090?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15815580#comment-15815580
]
nirav patel commented on SPARK-19090:
-------------------------------------
I tried passing spark parameters via oozie directly using <spark-opts>--conf
spark.executor.cores=5</spark-opts> which passes those parameters to :
org.apache.spark.deploy.SparkSubmit and then SparkSubmit subsequently calls my
application class. In this scenario I can see that dynamic executor scheduling
picking up this value and using it to request 5 vcores per executor. SO I think
that's my workaround.
Real issue seems that spark dynamic scheduling module is ignoring sparkConf
(spark.executor.cores) parameter set by user application class. It's
recognizing all other parameters! Because I am not passing any other parameter
directly to spark submit those are all set via my application code as shown in
my code snippet. It's only ignoring spark.executor.cores set at application
level which is weird. If I have read document correctly one can always override
spark command line parameter via application level SparkConf object. It
definitely works when dynamic scheduling is turned off . It only doesn't work
when its on.
> Dynamic Resource Allocation not respecting spark.executor.cores
> ---------------------------------------------------------------
>
> Key: SPARK-19090
> URL: https://issues.apache.org/jira/browse/SPARK-19090
> Project: Spark
> Issue Type: Bug
> Affects Versions: 1.5.2, 1.6.1, 2.0.1
> Reporter: nirav patel
>
> When enabling dynamic scheduling with yarn I see that all executors are using
> only 1 core even if I specify "spark.executor.cores" to 6. If dynamic
> scheduling is disabled then each executors will have 6 cores. i.e. it
> respects "spark.executor.cores". I have tested this against spark 1.5 . I
> think it will be the same behavior with 2.x as well.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
