You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "dzcxzl (Jira)" <ji...@apache.org> on 2020/10/14 11:28:00 UTC
[jira] [Created] (SPARK-33147) Avoid distribute user jar from
driver in yarn client mode
dzcxzl created SPARK-33147:
------------------------------
Summary: Avoid distribute user jar from driver in yarn client mode
Key: SPARK-33147
URL: https://issues.apache.org/jira/browse/SPARK-33147
Project: Spark
Issue Type: Improvement
Components: Spark Core
Affects Versions: 3.0.1
Reporter: dzcxzl
{code:java}
spark-submit --master yarn --deploy-mode client --num-executors 100 big_jar.jar
{code}
When the number of applied executors is large and the jar size is large, the executor pulls the jar from the driver, and the driver network traffic is high, and a timeout may occur. The driver and the executor of the yarn cluster may not be in the same data center.
{code:java}
20/10/04 00:46:02,269 [rpc-server-3-13] ERROR TransportRequestHandler: Error sending result StreamResponse{streamId=/jars/xxx-jar-with-dependencies.jar, byteCount=145417300, body=FileSegmentManagedBuffer{file=xxx-jar-with-dependencies.jar, offset=0, length=145417300}} to /x.x.x.x:33527; closing connection
{code}
We can automatically add the user's jar to --jars, which can avoid this problem.
--
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