1. 设置SCALA_VERSION
2. 执行conf/spark-env.sh
3. 设置CLASSPATH=<conf目录>
4. 如果存在assembly/target/scala-$SCALA_VERSION/spark-assembly*hadoop*-deps.jar，则添加
[core|repl|mllib|bagel|graphx|streaming]/target/scala-$SCALA_VERSION/classes:/assembly/target/scala-$SCALA_VERSION/spark-assembly*hadoop*-deps.jar
如果不存在，则检测RELEASE目录，存在则添加jars/spark-assembly*.jar，不存在则添加assembly/target/scala-$SCALA_VERSION/spark-assembly*hadoop*.jar
5. 检测SPARK_TESTING，为1则添加
[core|repl|mllib|bagel|graphx|streaming]/target/scala-$SCALA_VERSION/test-classes
6. 添加HADOOP_CONF_DIR、YARN_CONF_DIR