spark-submit参数说明--on YARN

示例： spark-submit [--option value] <application jar> [application arguments]

参数名称	含义
--master MASTER_URL	yarn
--deploy-mode DEPLOY_MODE	Driver程序运行的地方：client、cluster
--class CLASS_NAME	The FQCN of the class containing the main method of the application. For example, org.apache.spark.examples.SparkPi. 应用程序主类名称，含包名
--name NAME	应用程序名称
--jars JARS	Driver和Executor依赖的第三方jar包
--properties-file FILE	应用程序属性的文件路径，默认是conf/spark-defaults.conf
以下设置Driver
--driver-cores NUM	Driver程序使用的CPU核数(只用于cluster)，默认为1
--driver-memory MEM	Driver程序使用内存大小
--driver-library-path	Driver程序的库路径
--driver-class-path	Driver程序的类路径
--driver-java-options
以下设置Executor
--num-executors NUM	The total number of YARN containers to allocate for this application. Alternatively, you can use the spark.executor.instances configuration parameter. 启动的executor的数量，默认为2
--executor-cores NUM	Number of processor cores to allocate on each executor 每个executor使用的CPU核数，默认为1
--executor-memory MEM	The maximum heap size to allocate to each executor. Alternatively, you can use the spark.executor.memory configuration parameter. 每个executor内存大小，默认为1G
--queue QUEUE_NAME	The YARN queue to submit to. 提交应用程序给哪个YARN的队列，默认是default队列
--archives ARCHIVES
--files FILES	用逗号隔开的要放置在每个executor工作目录的文件列表

1.部署模式概述

   In YARN, each application instance has an ApplicationMaster process, which is the first container started for that application.

  The application is responsible for requesting resources from the ResourceManager, and, when allocated them, instructing NodeManagers to start containers on its behalf.

  ApplicationMasters obviate the need for an active client — the process starting the application can terminate and coordination continues from a process managed by YARN running on the cluster.

2.部署模式：Cluster

In cluster mode, the driver runs in the ApplicationMaster on a cluster host chosen by YARN.

This means that the same process, which runs in a YARN container, is responsible for both driving the application and requesting resources from YARN.

The client that launches the application doesn't need to continue running for the entire lifetime of the application.

Cluster mode is not well suited to using Spark interactively.

Spark applications that require user input, such as spark-shell and pyspark, need the Spark driver to run inside the client process that initiates the Spark application.

3.部署模式：Client

In client mode, the driver runs on the host where the job is submitted.

The ApplicationMaster is merely present to request executor containers from YARN.

The client communicates with those containers to schedule work after they start:

4.参考文档：

https://www.cloudera.com/documentation/enterprise/5-4-x/topics/cdh_ig_running_spark_on_yarn.html

http://spark.apache.org/docs/1.3.0/running-on-yarn.html