You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Cory Maklin (Jira)" <ji...@apache.org> on 2021/03/25 15:52:00 UTC
[jira] [Updated] (SPARK-34870) Jars downloaded with the --packages
argument are not added to the classpath for executors.
[ https://issues.apache.org/jira/browse/SPARK-34870?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Cory Maklin updated SPARK-34870:
--------------------------------
Description:
When Spark is run in local mode, it works as expected. However, when Spark is run in client mode, it copies the jars to the executor ($SPARK_HOME/work/<app id>/<executor id>), but never adds them to the classpath.
It might be worth noting that `spark.jars` does add the jars to the classpath, but unlike `spark.jars.packages` it doesn't automatically download the jar's compiled dependencies.
```
spark = SparkSession.builder\
.master(SPARK_MASTER)\
.appName(APP_NAME)\
...
.config("spark.jars.packages", DEPENDENCY_PACKAGES) \
...
.getOrCreate()
```
was:
When Spark is run in local mode, it works as expected. However, when Spark is run in client mode, it copies the jars to the executor ($SPARK_HOME/work/<app id>/<executor id>), but never adds them to the classpath.
It might be worth noting that `spark.jars` does add the jars to the classpath, but unlike `spark.jars.packages` it doesn't automatically download the jar's compiled dependencies.
```
spark = SparkSession.builder\
.master(SPARK_MASTER)\
.appName(APP_NAME)\
...
.config("spark.jars.packages", DEPENDENCY_PACKAGES) \
...
.getOrCreate()
```
> Jars downloaded with the --packages argument are not added to the classpath for executors.
> ------------------------------------------------------------------------------------------
>
> Key: SPARK-34870
> URL: https://issues.apache.org/jira/browse/SPARK-34870
> Project: Spark
> Issue Type: Bug
> Components: Spark Submit
> Affects Versions: 3.0.0
> Environment: Spark worker running inside a Kubernetes pod with a Bitnami Spark image, and the driver running inside of a Jupyter Spark Kubernetes pod.
> Reporter: Cory Maklin
> Priority: Major
>
> When Spark is run in local mode, it works as expected. However, when Spark is run in client mode, it copies the jars to the executor ($SPARK_HOME/work/<app id>/<executor id>), but never adds them to the classpath.
> It might be worth noting that `spark.jars` does add the jars to the classpath, but unlike `spark.jars.packages` it doesn't automatically download the jar's compiled dependencies.
>
> ```
> spark = SparkSession.builder\
> .master(SPARK_MASTER)\
> .appName(APP_NAME)\
> ...
> .config("spark.jars.packages", DEPENDENCY_PACKAGES) \
> ...
> .getOrCreate()
> ```
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org