问题描述
spark运行在yarn上。jdk:1.7.0_71hadoop:2.6spark:1.6spark配置:只配置了spark-env.sh文件,其他都没配置,配置如下exportSPARK_HOME=/yarn/spark-1.6.0-bin-hadoop2.6exportHADOOP_HOME=/yarn/hadoop-2.6.0exportYARN_CONF_DIR=/yarn/hadoop-2.6.0/etc/hadoopexportHBASE_CONF_DIR=/yarn/hbase-0.98.14-hadoop2/conf问题一:运行:一个简单的wordcount程序spark-submit--classcom.cyc.spark.JavaWordCount--masteryarn--deploy-modeclient/yarn/file/sparktest.jar一直打印:16/02/2321:11:30INFOClient:Applicationreportforapplication_1456228788292_0009(state:ACCEPTED)16/02/2321:11:31INFOClient:Applicationreportforapplication_1456228788292_0009(state:ACCEPTED)16/02/2321:11:32INFOClient:Applicationreportforapplication_1456228788292_0009(state:ACCEPTED)16/02/2321:11:33INFOClient:Applicationreportforapplication_1456228788292_0009(state:ACCEPTED)16/02/2321:11:34INFOClient:Applicationreportforapplication_1456228788292_0009(state:ACCEPTED)在yarn监管页面,显示求解,为什么程序不能跑?问题二:修改spark-env.sh文件,将exportYARN_CONF_DIR=/yarn/hadoop-2.6.0/etc/hadoop随便改一个路径,例如exportYARN_CONF_DIR=/yarn/hadoop程序也一直Applicationreportforapplication,在yarn的监管页面也有显示对应的application。求解,为什么路径不对也能被yarn接收到
解决方案
解决方案二:
楼主,问题一如何解决的呢?