[
https://issues.apache.org/jira/browse/SPARK-2454?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14069479#comment-14069479
]
Nan Zhu commented on SPARK-2454:
--------------------------------
there is a related issue and fix
https://issues.apache.org/jira/browse/SPARK-2404
https://github.com/apache/spark/pull/1331
where we should not overwrite SPARK_HOME in spark-submit and spark-class if the
user has set those two values
in our scenario, the remote cluster has the same user with the login portal,
say codingcat,
so the SPARK_HOME is set to /home/codingcat/spark-1.0, other users in the login
portal has a soft link to SPARK_HOME in their own directory,
however the current scripts overwrite the already-set SPARK_HOME to the pwd
running spark-submit, which does not exist in the cluster, causing the
exceptions like "cannot run /home/local_user/spark-class"....
> Separate driver spark home from executor spark home
> ---------------------------------------------------
>
> Key: SPARK-2454
> URL: https://issues.apache.org/jira/browse/SPARK-2454
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
> Affects Versions: 1.0.1
> Reporter: Andrew Or
> Fix For: 1.1.0
>
>
> The driver may not always share the same directory structure as the
> executors. It makes little sense to always re-use the driver's spark home on
> the executors.
> https://github.com/apache/spark/pull/1244/ is an open effort to fix this.
> However, this still requires us to set SPARK_HOME on all the executor nodes.
> Really we should separate this out into something like `spark.executor.home`
> and `spark.driver.home` rather than re-using SPARK_HOME everywhere.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
