简体   繁体   English

启动并提交工作火花

[英]launch and submit job spark

I try to "run" spark jobs width my java application, searching, I found the following two methods: ClientsArguments and SparkLauncher . 我尝试“运行”我的Java应用程序宽度的ClientsArguments作业,进行搜索,发现以下两种方法: ClientsArgumentsSparkLauncher

Could someone explain to me the difference between the two? 有人可以向我解释两者之间的区别吗? The difference between launch and submit job/application Spark? 启动和提交作业/应用程序之间的区别Spark? Thank you. 谢谢。

SparkLauncher is just a wrapper library for spark-submit , it coverts your SparkLauncher code to spark-submit script and then trigger the jobs. SparkLauncher只是用于spark-submit的包装库,它SparkLauncher您的SparkLauncher代码以执行spark-submit 脚本 ,然后触发作业。

The mechanism is the same with spark-submit script, if you look at the source code of SparkLauncher, it uses ProcessBuilder to construct the shell. 该机制与spark-submit脚本相同,如果您查看SparkLauncher的源代码,它使用ProcessBuilder构造外壳。

If you want to use SparkLauncher , you need to specify $JAVA_HOME , $SPRAK_HOME and other essential parameters. 如果要使用SparkLauncher ,则需要指定$JAVA_HOME$SPRAK_HOME和其他基本参数。 There are some limitation for SparkLauncher , the machine your SparkLauncher runs must have a $JAVA_HOME and $SPARK_HOME (Spark Library) which is used for SparkLauncher locating the script and related dependencies. SparkLauncher有一些限制,您的SparkLauncher运行的机器必须具有$JAVA_HOME$SPARK_HOME (火花库),用于SparkLauncher定位脚本和相关的依赖项。 This is soft of impossible for some cloud environment like CloudFoundry etc. 对于某些云环境(如CloudFoundry等)来说,这是不可能的。

You could assume that SparkLauncher equals spark-submit script, you could choose client or master, local or yarn mode. 您可以假设SparkLauncher等于spark-submit脚本,可以选择客户端或主模式,本地模式或yarn模式。

ClientsArguments is a class for YARN script, which is only works for yarn-mode. ClientsArguments是YARN脚本的类,仅适用于yarn-mode。

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM