sparkR在spark on yarn下的问题

sparkR在spark on yarn下的问题
官网上可以看到说明“Currently, SparkR supports running on YARN with the yarn-client mode. 

These steps show how to build SparkR with YARN support and run SparkR programs on a YARN cluster:”

具体见：

README.md

https://github.com/amplab-extras/SparkR-pkg

但是也看到工作图，里面会涉及到每个节点，Spark Executor和RScript的交互。

插图：

之前测试的sparkR只涉及到从hive获取数据源，没有问题，但是没有测试，执行R脚本的情形。

测试的结果是spark on yarn的时候，需要跑R脚本的情形下，发现抛出“RScript”找不到的情形，开始怀疑是R安装和环境变量PATH的问题。
但是不用修改任何东西，用standalone方式提交，马上正常。Stackoverflow上看到的情形也多是standalone的情况。
所以这种情况下，为了正常使用sparkR，还是直接standalone的方式。
#################################
yarn:

#################################

sparkR --executor-memory 2g --driver-cores 2 --master yarn --num-executors 3
插图：

########################
Standalone:

########################

sparkR --executor-memory 2g --total-executor-cores 10 --master spark://master.hadoop:7077
插图：

############
#环境变量
.libPaths(c(.libPaths(), '/opt/hadoop/spark-latest/R/lib')) 
Sys.setenv(SPARK_HOME = '/opt/hadoop/spark-latest') 
Sys.setenv(PATH = paste(Sys.getenv(c('PATH')), '/opt/hadoop/spark-latest/bin', sep=':')) 
Sys.setenv(HADOOP_CMD="/opt/cloudera/parcels/CDH/bin/hadoop")
Sys.setenv(HADOOP_HOME="/opt/cloudera/parcels/CDH/lib/hadoop")
Sys.setenv(HADOOP_CONF_DIR="/etc/hadoop/conf")
Sys.setenv(YARN_CONF_DIR="/etc/hadoop/conf")
Sys.setenv(HIVE_HOME="/opt/cloudera/parcels/CDH/lib/hive")
Sys.setenv(SCALA_HOME="/opt/hadoop/scala-latest")
#加载包
library("rJava")  
library("rhdfs")
library("SparkR")
#初始化
#sc <- sparkR.init("yarn-client", "SparkR", "/opt/hadoop/spark-latest",list(spark.executor.memory="1g"))

rdd <- SparkR:::textFile(sc, '/user/hive/warehouse/pcp2/city_test.txt')
counts <- SparkR:::map(rdd, nchar)
SparkR:::take(counts, 3)

hdfs.init()
hdfs.cat("/user/hive/warehouse/pcp2/city_test.txt")


####################################### 附上帮助
[hadoop@snn conf]$ sparkR --help
Usage: ./bin/sparkR [options]

Options:
  --master MASTER_URL         spark://host:port, mesos://host:port, yarn, or local.
  --deploy-mode DEPLOY_MODE   Whether to launch the driver program locally ("client") or
                              on one of the worker machines inside the cluster ("cluster")
                              (Default: client).
  --class CLASS_NAME          Your application's main class (for Java / Scala apps).
  --name NAME                 A name of your application.
  --jars JARS                 Comma-separated list of local jars to include on the driver
                              and executor classpaths.
  --packages                  Comma-separated list of maven coordinates of jars to include
                              on the driver and executor classpaths. Will search the local
                              maven repo, then maven central and any additional remote
                              repositories given by --repositories. The format for the
                              coordinates should be groupId:artifactId:version.
  --exclude-packages          Comma-separated list of groupId:artifactId, to exclude while
                              resolving the dependencies provided in --packages to avoid
                              dependency conflicts.
  --repositories              Comma-separated list of additional remote repositories to
                              search for the maven coordinates given with --packages.
  --py-files PY_FILES         Comma-separated list of .zip, .egg, or .py files to place
                              on the PYTHONPATH for Python apps.
  --files FILES               Comma-separated list of files to be placed in the working
                              directory of each executor.


  --conf PROP=VALUE           Arbitrary Spark configuration property.
  --properties-file FILE      Path to a file from which to load extra properties. If not
                              specified, this will look for conf/spark-defaults.conf.


  --driver-memory MEM         Memory for driver (e.g. 1000M, 2G) (Default: 1024M).
  --driver-java-options       Extra Java options to pass to the driver.
  --driver-library-path       Extra library path entries to pass to the driver.
  --driver-class-path         Extra class path entries to pass to the driver. Note that
                              jars added with --jars are automatically included in the
                              classpath.


  --executor-memory MEM       Memory per executor (e.g. 1000M, 2G) (Default: 1G).


  --proxy-user NAME           User to impersonate when submitting the application.


  --help, -h                  Show this help message and exit
  --verbose, -v               Print additional debug output
  --version,                  Print the version of current Spark


 Spark standalone with cluster deploy mode only:
  --driver-cores NUM          Cores for driver (Default: 1).


 Spark standalone or Mesos with cluster deploy mode only:
  --supervise                 If given, restarts the driver on failure.
  --kill SUBMISSION_ID        If given, kills the driver specified.
  --status SUBMISSION_ID      If given, requests the status of the driver specified.


 Spark standalone and Mesos only:
  --total-executor-cores NUM  Total cores for all executors.


 Spark standalone and YARN only:
  --executor-cores NUM        Number of cores per executor. (Default: 1 in YARN mode,
                              or all available cores on the worker in standalone mode)


 YARN-only:
  --driver-cores NUM          Number of cores used by the driver, only in cluster mode
                              (Default: 1).
  --queue QUEUE_NAME          The YARN queue to submit to (Default: "default").
  --num-executors NUM         Number of executors to launch (Default: 2).
  --archives ARCHIVES         Comma separated list of archives to be extracted into the
                              working directory of each executor.
  --principal PRINCIPAL       Principal to be used to login to KDC, while running on
                              secure HDFS.
  --keytab KEYTAB             The full path to the file that contains the keytab for the
                              principal specified above. This keytab will be copied to
                              the node running the Application Master via the Secure
                              Distributed Cache, for renewing the login tickets and the
                              delegation tokens periodically.


[hadoop@snn conf]$