Merge pull request apache#345 from colorant/yarn

support distributing extra files to worker for yarn client mode

So that user doesn't need to package all dependency into one assemble jar as spark app jar

Author: tgravescs

docs/running-on-yarn.md

@@ -114,6 +114,8 @@ For example:
     SPARK_YARN_APP_JAR=examples/target/scala-{{site.SCALA_VERSION}}/spark-examples-assembly-{{site.SPARK_VERSION}}.jar \
     MASTER=yarn-client ./bin/spark-shell
 
+You can also send extra files to yarn cluster for worker to use by exporting SPARK_YARN_DIST_FILES=file1,file2... etc.
+
 # Building Spark for Hadoop/YARN 2.2.x
 
 See [Building Spark with Maven](building-with-maven.html) for instructions on how to build Spark using the Maven process.

yarn/alpha/src/main/scala/org/apache/spark/deploy/yarn/Client.scala

@@ -102,7 +102,7 @@ class Client(args: ClientArguments, conf: Configuration, sparkConf: SparkConf)
       (System.getenv("SPARK_JAR") == null) -> "Error: You must set SPARK_JAR environment variable!",
       (args.userJar == null) -> "Error: You must specify a user jar!",
       (args.userClass == null) -> "Error: You must specify a user class!",
-      (args.numWorkers <= 0) -> "Error: You must specify atleast 1 worker!",
+      (args.numWorkers <= 0) -> "Error: You must specify at least 1 worker!",
       (args.amMemory <= YarnAllocationHandler.MEMORY_OVERHEAD) -> ("Error: AM memory size must be " +
         "greater than: " + YarnAllocationHandler.MEMORY_OVERHEAD),
       (args.workerMemory <= YarnAllocationHandler.MEMORY_OVERHEAD) -> ("Error: Worker memory size " +

yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClientSchedulerBackend.scala

@@ -39,6 +39,7 @@ private[spark] class YarnClientSchedulerBackend(
     val defaultWorkerNumber = "1"
 
     val userJar = System.getenv("SPARK_YARN_APP_JAR")
+    val distFiles = System.getenv("SPARK_YARN_DIST_FILES")
     var workerCores = System.getenv("SPARK_WORKER_CORES")
     var workerMemory = System.getenv("SPARK_WORKER_MEMORY")
     var workerNumber = System.getenv("SPARK_WORKER_INSTANCES")
@@ -64,7 +65,8 @@ private[spark] class YarnClientSchedulerBackend(
       "--worker-memory", workerMemory,
       "--worker-cores", workerCores,
       "--num-workers", workerNumber,
-      "--master-class", "org.apache.spark.deploy.yarn.WorkerLauncher"
+      "--master-class", "org.apache.spark.deploy.yarn.WorkerLauncher",
+      "--files", distFiles
     )
 
     val args = new ClientArguments(argsArray, conf)

yarn/stable/src/main/scala/org/apache/spark/deploy/yarn/Client.scala

@@ -122,7 +122,7 @@ class Client(args: ClientArguments, conf: Configuration, sparkConf: SparkConf)
       (System.getenv("SPARK_JAR") == null) -> "Error: You must set SPARK_JAR environment variable!",
       (args.userJar == null) -> "Error: You must specify a user jar!",
       (args.userClass == null) -> "Error: You must specify a user class!",
-      (args.numWorkers <= 0) -> "Error: You must specify atleast 1 worker!",
+      (args.numWorkers <= 0) -> "Error: You must specify at least 1 worker!",
       (args.amMemory <= YarnAllocationHandler.MEMORY_OVERHEAD) -> ("Error: AM memory size must be" +
         "greater than: " + YarnAllocationHandler.MEMORY_OVERHEAD),
       (args.workerMemory <= YarnAllocationHandler.MEMORY_OVERHEAD) -> ("Error: Worker memory size" +