spark shell 删除失效_Spark任务提交源码解析

1. 前言

反反复复捣鼓了很久，终于开始学习Spark的源码了，果不其然，那真的很有趣。这里我打算一本正经的胡说八道来讲一下Spark作业的提交过程。

基础mac系统基础环境如下：

JDK 1.8
IDEA 2019.3
源码Spark 2.3.3
Scala 2.11.8
提交脚本
# 事先准备好的Spark任务(源码example LocalPi)基于local模式
bash spark-submit
--class com.lp.test.app.LocalPi
--master local
/Users/lipan/Desktop/spark-local/original-spark-local-train-1.0.jar
10

2. 提交流程

我们在提交Spark任务时都是从spark-submit(或者spark-shell)来提交一个作业的，从spark-submit脚本一步步深入进去看看任务的整体提交流程。首先看一下整体的流程概要图：

根据上图中的整体流程，接下来我们对里面的每一个流程的源码进行一一剖析跟踪。

2.1 spark-submit脚本

#!/usr/bin/env bash## 如果SPARK_HOME变量没有设置值，则执行当前目录下的find-spark-home脚本文件，设置SPARK_HOME值if [ -z "${SPARK_HOME}" ]; then  source "$(dirname "$0")"/find-spark-homefiecho "${SPARK_HOME}"# disable randomized hash for string in Python 3.3+export PYTHONHASHSEED=0# 这里可以看到将接收到的参数提交到了spark-class脚本执行exec "${SPARK_HOME}"/bin/spark-class org.apache.spark.deploy.SparkSubmit "$@"

2.2 spark-class脚本

#!/usr/bin/env bashif [ -z "${SPARK_HOME}" ]; then  source "$(dirname "$0")"/find-spark-homefi# 配置一些环境变量,它会将conf/spark-env.sh中的环境变量加载进来:. "${SPARK_HOME}"/bin/load-spark-env.sh# Find the java binary  如果有java_home环境变量会将java_home/bin/java给RUNNERif [ -n "${JAVA_HOME}" ]; then  RUNNER="${JAVA_HOME}/bin/java"else  if [ "$(command -v java)" ]; then    RUNNER="java"  else    echo "JAVA_HOME is not set" >&2    exit 1  fifi# Find Spark jars.# 这一段，主要是寻找java命令 寻找spark的jar包# 这里如果我们的jar包数量多，而且内容大，可以事先放到每个机器的对应目录下，这里是一个优化点if [ -d "${SPARK_HOME}/jars" ]; then  SPARK_JARS_DIR="${SPARK_HOME}/jars"else  SPARK_JARS_DIR="${SPARK_HOME}/assembly/target/scala-$SPARK_SCALA_VERSION/jars"fiif [ ! -d "$SPARK_JARS_DIR" ] && [ -z "$SPARK_TESTING$SPARK_SQL_TESTING" ]; then  echo "Failed to find Spark jars directory ($SPARK_JARS_DIR)." 1>&2  echo "You need to build Spark with the target "package" before running this program." 1>&2  exit 1else  LAUNCH_CLASSPATH="$SPARK_JARS_DIR/*"fi# Add the launcher build dir to the classpath if requested.if [ -n "$SPARK_PREPEND_CLASSES" ]; then  LAUNCH_CLASSPATH="${SPARK_HOME}/launcher/target/scala-$SPARK_SCALA_VERSION/classes:$LAUNCH_CLASSPATH"fi# For testsif [[ -n "$SPARK_TESTING" ]]; then  unset YARN_CONF_DIR  unset HADOOP_CONF_DIRfi# The launcher library will print arguments separated by a NULL character, to allow arguments with# characters that would be otherwise interpreted by the shell. Read that in a while loop, populating# an array that will be used to exec the final command.# 启动程序库将打印由NULL字符分隔的参数，以允许与shell进行其他解释的字符进行参数。在while循环中读取它，填充将用于执行最终命令的数组。## The exit code of the launcher is appended to the output, so the parent shell removes it from the# command array and checks the value to see if the launcher succeeded.# 启动程序的退出代码被追加到输出，因此父shell从命令数组中删除它，并检查其值，看看启动器是否成功。# 这里spark启动了以SparkSubmit为主类的JVM进程。build_command() {  "$RUNNER" -Xmx128m -cp "$LAUNCH_CLASSPATH" org.apache.spark.launcher.Main "$@"  printf "%d0" $?}# Turn off posix mode since it does not allow process substitution# 关闭posix模式，因为它不允许进程替换。# 调用build_command org.apache.spark.launcher.Main拼接提交命令set +o posixCMD=()while IFS= read -d '' -r ARG; do  CMD+=("$ARG")done < &2  exit 1fiif [ $LAUNCHER_EXIT_CODE != 0 ]; then  exit $LAUNCHER_EXIT_CODEfiCMD=("${CMD[@]:0:$LAST}")# ${CMD[@]} 参数如下# /Library/Java/JavaVirtualMachines/jdk1.8.0_172.jdk/Contents/Home/bin/java -cp /Users/lipan/workspace/source_code/spark-2.3.3/conf/:/Users/lipan/workspace/source_code/spark-2.3.3/assembly/target/scala-2.11/jars/* -Xmx1g org.apache.spark.deploy.SparkSubmit --master local --class com.lp.test.app.LocalPi /Users/lipan/Desktop/spark-local/original-spark-local-train-1.0.jar 10exec "${CMD[@]}"

相对于spark-submit，spark-class文件的执行逻辑稍显复杂，总体如下：

检查SPARK_HOME执行环境
执行load-spark-env.sh文件，加载一些默认的环境变量(包括加载spark-env.sh文件)
检查JAVA_HOME执行环境
寻找Spark相关的jar包
执行org.apache.spark.launcher.Main解析参数，构建CMD命令
CMD命令判断
执行org.apache.spark.deploy.SparkSubmit这个类。

2.3 org.apache.spark.launcher.Main

java -Xmx128m -cp ...jars org.apache.spark.launcher.Main "$@"

也就是说org.apache.spark.launcher.Main是被spark-class调用，从spark-class接收参数。这个类是提供spark内部脚本调用的工具类，并不是真正的执行入口。它负责调用其他类，对参数进行解析，并生成执行命令，最后将命令返回给spark-class的 exec “${CMD[@]}”执行。

可以把”$@”执行相关参数带入IDEA中的org.apache.spark.launcher.Main方法中执行，操作参考如下：

package org.apache.spark.launcher;import java.util.ArrayList;import java.util.Arrays;import java.util.HashMap;import java.util.List;import java.util.Map;import static org.apache.spark.launcher.CommandBuilderUtils.*;/** * Command line interface for the Spark launcher. Used internally by Spark scripts. * 这是提供spark内部脚本使用工具类 */ class Main {    /**     * Usage: Main [class] [class args]     * 分为spark-submit和spark-class两种模式     * 如果提交的是class类的话,会包含其他如:master/worker/history等等     * unix系统的输出的参数是集合,而windows参数是空格分隔     *     * spark-class提交过来的参数如下：     * org.apache.spark.deploy.SparkSubmit      * --class com.lp.test.app.LocalPi      * --master local      * /Users/lipan/Desktop/spark-local/spark-local-train-1.0.jar     */    public static void main(String[] argsArray) throws Exception {        checkArgument(argsArray.length > 0, "Not enough arguments: missing class name.");        // 判断参数列表        List args = new ArrayList<>(Arrays.asList(argsArray));        String className = args.remove(0);        // 判断是否打印执行信息        boolean printLaunchCommand = !isEmpty(System.getenv("SPARK_PRINT_LAUNCH_COMMAND"));        // 创建命令解析器        AbstractCommandBuilder builder;        /**         * 构建执行程序对象:spark-submit/spark-class         * 把参数都取出并解析,放入执行程序对象中         * 意思是,submit还是master和worker等程序在这里拆分,并获取对应的执行参数         */        if (className.equals("org.apache.spark.deploy.SparkSubmit")) {            try {                // 构建spark-submit命令对象                builder = new SparkSubmitCommandBuilder(args);            } catch (IllegalArgumentException e) {                printLaunchCommand = false;                System.err.println("Error: " + e.getMessage());                System.err.println();                // 类名解析--class org.apache.spark.repl.Main                MainClassOptionParser parser = new MainClassOptionParser();                try {                    parser.parse(args);                } catch (Exception ignored) {                    // Ignore parsing exceptions.                }                // 帮助信息                List help = new ArrayList<>();                if (parser.className != null) {                    help.add(parser.CLASS);                    help.add(parser.className);                }                help.add(parser.USAGE_ERROR);                // 构建spark-submit帮助信息对象                builder = new SparkSubmitCommandBuilder(help);            }        } else {            // 构建spark-class命令对象            // 主要是在这个类里解析了命令对象和参数            builder = new SparkClassCommandBuilder(className, args);        }        /**         * 这里才真正构建了执行命令         * 调用了SparkClassCommandBuilder的buildCommand方法         * 把执行参数解析成了k/v格式         */        Map env = new HashMap<>();        List cmd = builder.buildCommand(env);        if (printLaunchCommand) {            System.err.println("Spark Command: " + join(" ", cmd));            System.err.println("========================================");        }        if (isWindows()) {            System.out.println(prepareWindowsCommand(cmd, env));        } else {            // In bash, use NULL as the arg separator since it cannot be used in an argument.            /**             * 输出参数：/Library/Java/JavaVirtualMachines/jdk1.8.0_172.jdk/Contents/Home/bin/java             * -cp /Users/lipan/workspace/source_code/spark-2.3.3/conf/:/Users/lipan/workspace/source_code/spark-2.3.3/assembly/target/scala-2.11/jars/*             * -Xmx1g org.apache.spark.deploy.SparkSubmit             * --master local             * --class com.lp.test.app.LocalPi             * /Users/lipan/Desktop/spark-local/original-spark-local-train-1.0.jar 10             *  java -cp / org.apache.spark.deploy.SparkSubmit启动该类             */            List bashCmd = prepareBashCommand(cmd, env);            for (String c : bashCmd) {                System.out.print(c);                System.out.print('0');            }        }    }    /**     * windows环境下     */    private static String prepareWindowsCommand(List cmd, Map childEnv) {        StringBuilder cmdline = new StringBuilder();        for (Map.Entry e : childEnv.entrySet()) {            cmdline.append(String.format("set %s=%s", e.getKey(), e.getValue()));            cmdline.append(" && ");        }        for (String arg : cmd) {            cmdline.append(quoteForBatchScript(arg));            cmdline.append(" ");        }        return cmdline.toString();    }    /**     * bash环境，如：Linux     */    private static List prepareBashCommand(List cmd, Map childEnv) {        if (childEnv.isEmpty()) {            return cmd;        }        List newCmd = new ArrayList<>();        newCmd.add("env");        for (Map.Entry e : childEnv.entrySet()) {            newCmd.add(String.format("%s=%s", e.getKey(), e.getValue()));        }        newCmd.addAll(cmd);        return newCmd;    }    /**     * 当spark-submit提交失败时,这里会再进行一次解析,再不行才会提示用法     */    private static class MainClassOptionParser extends SparkSubmitOptionParser {        String className;        @Override        protected boolean handle(String opt, String value) {            if (CLASS.equals(opt)) {                className = value;            }            return false;        }        @Override        protected boolean handleUnknown(String opt) {            return false;        }        @Override        protected void handleExtraArgs(List extra) {        }    }   }

Main中主要涉及到的一些类SparkSubmitCommandBuilder、SparkClassCommandBuilder 和 buildCommand都是在对参数和构建命令进行处理，这里不一一展开详解。

2.4 org.apache.spark.deploy.SparkSubmit

org.apache.spark.launcher.Main中会解析过滤参数，构建执行命令，返回给spark-class脚本，最后通过 exec “${CMD[@]}” 真正调用SparkSubmit类。

可通过解析后提交的参数”$@”设置在IDEA中逐步跟踪源码，操作参考如下：

2.4.1 SparkSubmitAction

在org.apache.spark.launcher.Main类的最前面定义了一个类SparkSubmitAction枚举状态类。

/** * Whether to submit, kill, or request the status of an application. * The latter two operations are currently supported only for standalone and Mesos cluster modes. * 这个类主要是提交app,终止和请求状态,但目前终止和请求只能在standalone和mesos模式下 */// 继承了枚举类,定义了4个属性,多了一个打印spark版本private[deploy] object SparkSubmitAction extends Enumeration {  type SparkSubmitAction = Value  val SUBMIT, KILL, REQUEST_STATUS, PRINT_VERSION = Value}

2.4.2 SparkSubmit

在SparkSubmit类中的方法执行可参考如下，在每个方法中都有详细的注释。具体细节也可以根据文末的链接地址中载源码断进行断点调试。

2.4.2.1 Main

override def main(args: Array[String]): Unit = {    // 初始化logging系统,并跟日志判断是否需要在app启动时重启    val uninitLog = initializeLogIfNecessary(true, silent = true)    /**     * 构建spark提交需要的参数并进行赋值 SparkSubmitArguments     * 1.解析参数     * 2.从属性文件填充“sparkProperties”映射(未指定默认情况下未：spark-defaults.conf)     * 3.移除不是以"spark." 开头的变量     * 4.参数填充对应到实体属性上     * 5.action参数验证     */    val appArgs = new SparkSubmitArguments(args)    // 参数不重复则输出配置    if (appArgs.verbose) {      printStream.println(appArgs)    }    appArgs.action match {      case SparkSubmitAction.SUBMIT => submit(appArgs, uninitLog)      case SparkSubmitAction.KILL => kill(appArgs)      case SparkSubmitAction.REQUEST_STATUS => requestStatus(appArgs)    }  }

2.4.2.2 submit

  /**   * 通过匹配SUBMIT执行的submit()   *   * 首先是根据不同调度模式和yarn不同模式,导入调用类的路径,默认配置及输入参数,准备相应的启动环境   * 然后通过对应的环境来调用相应子类的main方法   * 这里因为涉及到重复调用,所以采用了@tailrec尾递归,即重复调用方法的最后一句并返回结果   * 即:runMain(childArgs, childClasspath, sparkConf, childMainClass, args.verbose)   */  @tailrec  private def submit(args: SparkSubmitArguments, uninitLog: Boolean): Unit = {    /**     * 先准备运行环境,传入解析的各种参数     * 这里会先进入     * lazy val secMgr = new SecurityManager(sparkConf)     * 先初始化SecurityManager后,再进入prepareSubmitEnvironment()     * prepareSubmitEnvironment()代码比较长,放到最下面去解析     */    val (childArgs, childClasspath, sparkConf, childMainClass) = prepareSubmitEnvironment(args)    // 主要是调用runMain()启动相应环境的main()的方法    // 环境准备好以后,会先往下运行判断,这里是在等着调用    def doRunMain(): Unit = {      // 提交时可以指定--proxy-user,如果没有指定,则获取当前用户      if (args.proxyUser != null) {        val proxyUser = UserGroupInformation.createProxyUser(args.proxyUser,          UserGroupInformation.getCurrentUser())        try {          proxyUser.doAs(new PrivilegedExceptionAction[Unit]() {            override def run(): Unit = {              // 这里是真正的执行,runMain()              runMain(childArgs, childClasspath, sparkConf, childMainClass, args.verbose)            }          })        } catch {          case e: Exception =>            // Hadoop's AuthorizationException suppresses the exception's stack trace, which            // makes the message printed to the output by the JVM not very helpful. Instead,            // detect exceptions with empty stack traces here, and treat them differently.            if (e.getStackTrace().length == 0) {              // scalastyle:off println              printStream.println(s"ERROR: ${e.getClass().getName()}: ${e.getMessage()}")              // scalastyle:on println              exitFn(1)            } else {              throw e            }        }      } else {        // 没有指定用户时执行        runMain(childArgs, childClasspath, sparkConf, childMainClass, args.verbose)      }    }    // 启动main后重新初始化logging    if (uninitLog) {      Logging.uninitialize()    }    // standalone模式有两种提交网关,    // (1)使用o.a.s.apply.client作为包装器的传统RPC网关和基于REST服务的网关    // (2)spark1.3后默认使用REST    // 如果master终端没有使用REST服务,spark会故障切换到RPC 这里判断standalone模式和使用REST服务    if (args.isStandaloneCluster && args.useRest) {      // 异常捕获,判断正确的话输出信息,进入doRunMain()      try {        logInfo("Running Spark using the REST application submission protocol.")        doRunMain()      } catch {        // Fail over to use the legacy submission gateway        // 否则异常输出信息,并设置submit失败        case e: SubmitRestConnectionException =>          logWarning(s"Master endpoint ${args.master} was not a REST server. " +            "Falling back to legacy submission gateway instead.")          args.useRest = false          submit(args, false)      }      // In all other modes, just run the main class as prepared      // 其他模式,按准备的环境调用上面的doRunMain()运行相应的main()      // 在进入前,初始化了SparkContext和SparkSession    } else {      doRunMain()    }  }

2.4.2.3 prepareSubmitEnvironment

  /**   * 准备各种模式的配置参数   *   * @param args 用于环境准备的已分析SparkSubmitArguments   * @param conf 在Hadoop配置中，仅在单元测试中设置此参数。   * @return a 4-tuple:   *         (1) the arguments for the child process,   *         (2) a list of classpath entries for the child,   *         (3) a map of system properties, and   *         (4) the main class for the child   *         返回一个4元组(childArgs, childClasspath, sparkConf, childMainClass)   *         childArgs：子进程的参数   *         childClasspath：子级的类路径条目列表   *         sparkConf：系统参数map集合   *         childMainClass：子级的主类   *   *         Exposed for testing.   *   *         由于不同的部署方式其卖弄函数是不一样的，主要是由spark的提交参数决定   */  private[deploy] def prepareSubmitEnvironment(                                                args: SparkSubmitArguments,                                                conf: Option[HadoopConfiguration] = None)  : (Seq[String], Seq[String], SparkConf, String) = {    try {      doPrepareSubmitEnvironment(args, conf)    } catch {      case e: SparkException =>        printErrorAndExit(e.getMessage)        throw e    }  }    private def doPrepareSubmitEnvironment(                                          args: SparkSubmitArguments,                                          conf: Option[HadoopConfiguration] = None)  : (Seq[String], Seq[String], SparkConf, String) = {    // Return values    val childArgs = new ArrayBuffer[String]()    val childClasspath = new ArrayBuffer[String]()    // SparkConf 会默认加一些系统参数    val sparkConf = new SparkConf()    var childMainClass = ""    // 设置集群模式    // 也就是提交时指定--master local/yarn/yarn-client/yarn-cluster/spark://192.168.2.1:7077或者 mesos,k8s等运行模式    val clusterManager: Int = args.master match {      case "yarn" => YARN      case "yarn-client" | "yarn-cluster" =>        printWarning(s"Master ${args.master} is deprecated since 2.0." +          " Please use master "yarn" with specified deploy mode instead.")        YARN      case m if m.startsWith("spark") => STANDALONE      case m if m.startsWith("mesos") => MESOS      case m if m.startsWith("k8s") => KUBERNETES      case m if m.startsWith("local") => LOCAL      case _ =>        printErrorAndExit("Master must either be yarn or start with spark, mesos, k8s, or local")        -1    }    // 设置部署模式 --deploy-mode    var deployMode: Int = args.deployMode match {      case "client" | null => CLIENT      case "cluster" => CLUSTER      case _ => printErrorAndExit("Deploy mode must be either client or cluster"); -1    }    //由于指定“yarn-cluster”和“yarn-client”的不受支持的方式封装了主模式和部署模式，    // 因此我们有一些逻辑来推断master和部署模式(如果只指定一种模式)，或者在它们不一致时提前退出    if (clusterManager == YARN) {      (args.master, args.deployMode) match {        case ("yarn-cluster", null) =>          deployMode = CLUSTER          args.master = "yarn"        case ("yarn-cluster", "client") =>          printErrorAndExit("Client deploy mode is not compatible with master "yarn-cluster"")        case ("yarn-client", "cluster") =>          printErrorAndExit("Cluster deploy mode is not compatible with master "yarn-client"")        case (_, mode) =>          args.master = "yarn"      }      // Make sure YARN is included in our build if we're trying to use it      if (!Utils.classIsLoadable(YARN_CLUSTER_SUBMIT_CLASS) && !Utils.isTesting) {        printErrorAndExit(          "Could not load YARN classes. " +            "This copy of Spark may not have been compiled with YARN support.")      }    }    // 判断k8s模式master和非testing模式    if (clusterManager == KUBERNETES) {      args.master = Utils.checkAndGetK8sMasterUrl(args.master)      // Make sure KUBERNETES is included in our build if we're trying to use it      if (!Utils.classIsLoadable(KUBERNETES_CLUSTER_SUBMIT_CLASS) && !Utils.isTesting) {        printErrorAndExit(          "Could not load KUBERNETES classes. " +            "This copy of Spark may not have been compiled with KUBERNETES support.")      }    }    // 错判断不可用模式    (clusterManager, deployMode) match {      case (STANDALONE, CLUSTER) if args.isPython =>        printErrorAndExit("Cluster deploy mode is currently not supported for python " +          "applications on standalone clusters.")      case (STANDALONE, CLUSTER) if args.isR =>        printErrorAndExit("Cluster deploy mode is currently not supported for R " +          "applications on standalone clusters.")      case (KUBERNETES, _) if args.isPython =>        printErrorAndExit("Python applications are currently not supported for Kubernetes.")      case (KUBERNETES, _) if args.isR =>        printErrorAndExit("R applications are currently not supported for Kubernetes.")      case (KUBERNETES, CLIENT) =>        printErrorAndExit("Client mode is currently not supported for Kubernetes.")      case (LOCAL, CLUSTER) =>        printErrorAndExit("Cluster deploy mode is not compatible with master "local"")      case (_, CLUSTER) if isShell(args.primaryResource) =>        printErrorAndExit("Cluster deploy mode is not applicable to Spark shells.")      case (_, CLUSTER) if isSqlShell(args.mainClass) =>        printErrorAndExit("Cluster deploy mode is not applicable to Spark SQL shell.")      case (_, CLUSTER) if isThriftServer(args.mainClass) =>        printErrorAndExit("Cluster deploy mode is not applicable to Spark Thrift server.")      case _ =>    }    // args.deployMode为空则设置deployMode值为参数,因为上面判断了args.deployMode为空deployMode为client    (args.deployMode, deployMode) match {      case (null, CLIENT) => args.deployMode = "client"      case (null, CLUSTER) => args.deployMode = "cluster"      case _ =>    }    // 根据资源管理器和部署模式，进行逻辑判断出几种特殊运行方式。    val isYarnCluster = clusterManager == YARN && deployMode == CLUSTER    val isMesosCluster = clusterManager == MESOS && deployMode == CLUSTER    val isStandAloneCluster = clusterManager == STANDALONE && deployMode == CLUSTER    val isKubernetesCluster = clusterManager == KUBERNETES && deployMode == CLUSTER    // 这里主要是添加相关的依赖    if (!isMesosCluster && !isStandAloneCluster) {      // 如果有maven依赖项，则解析它们，并将类路径添加到jar中。对于包含Python代码的包，也将它们添加到py文件中      val resolvedMavenCoordinates = DependencyUtils.resolveMavenDependencies(        args.packagesExclusions, args.packages, args.repositories, args.ivyRepoPath,        args.ivySettingsPath)      if (!StringUtils.isBlank(resolvedMavenCoordinates)) {        args.jars = mergeFileLists(args.jars, resolvedMavenCoordinates)        if (args.isPython || isInternal(args.primaryResource)) {          args.pyFiles = mergeFileLists(args.pyFiles, resolvedMavenCoordinates)        }      }      // 安装任何可能通过--jar或--packages传递的R包。Spark包可能在jar中包含R源代码。      if (args.isR && !StringUtils.isBlank(args.jars)) {        RPackageUtils.checkAndBuildRPackage(args.jars, printStream, args.verbose)      }    }    args.sparkProperties.foreach { case (k, v) => sparkConf.set(k, v) }    // sparkConf 加载Hadoop相关配置文件    val hadoopConf = conf.getOrElse(SparkHadoopUtil.newConfiguration(sparkConf))    // 工作临时目录    val targetDir = Utils.createTempDir()    //  判断当前模式下sparkConf的k/v键值对中key是否在JVM中全局可用    // 确保keytab在JVM中的任何位置都可用(keytab是Kerberos的身份认证,详情可参考：http://ftuto.lofter.com/post/31e97f_6ad659f)    if (clusterManager == YARN || clusterManager == LOCAL || clusterManager == MESOS) {      // 当前运行环境的用户不为空,args中yarn模式参数key列表不为空,则提示key列表文件不存在      if (args.principal != null) {        if (args.keytab != null) {          require(new File(args.keytab).exists(), s"Keytab file: ${args.keytab} does not exist")          // 在sysProps中添加keytab和主体配置，以供以后使用;例如，在spark sql中，用于与HiveMetastore对话的隔离类装入器将使用这些设置。          // 它们将被设置为Java系统属性，然后由SparkConf加载          sparkConf.set(KEYTAB, args.keytab)          sparkConf.set(PRINCIPAL, args.principal)          UserGroupInformation.loginUserFromKeytab(args.principal, args.keytab)        }      }    }    // Resolve glob path for different resources.    // 设置全局资源,也就是合并各种模式依赖的路径的资源和hadoopConf中设置路径的资源，各种jars,file,pyfile和压缩包    args.jars = Option(args.jars).map(resolveGlobPaths(_, hadoopConf)).orNull    args.files = Option(args.files).map(resolveGlobPaths(_, hadoopConf)).orNull    args.pyFiles = Option(args.pyFiles).map(resolveGlobPaths(_, hadoopConf)).orNull    args.archives = Option(args.archives).map(resolveGlobPaths(_, hadoopConf)).orNull    // 创建SecurityManager实例    lazy val secMgr = new SecurityManager(sparkConf)    // 在Client模式下，下载远程资源文件。    var localPrimaryResource: String = null    var localJars: String = null    var localPyFiles: String = null    if (deployMode == CLIENT) {      localPrimaryResource = Option(args.primaryResource).map {        downloadFile(_, targetDir, sparkConf, hadoopConf, secMgr)      }.orNull      localJars = Option(args.jars).map {        downloadFileList(_, targetDir, sparkConf, hadoopConf, secMgr)      }.orNull      localPyFiles = Option(args.pyFiles).map {        downloadFileList(_, targetDir, sparkConf, hadoopConf, secMgr)      }.orNull    }    // When running in YARN, for some remote resources with scheme:    //   1. Hadoop FileSystem doesn't support them.    //   2. We explicitly bypass Hadoop FileSystem with "spark.yarn.dist.forceDownloadSchemes".    // We will download them to local disk prior to add to YARN's distributed cache.    // For yarn client mode, since we already download them with above code, so we only need to    // figure out the local path and replace the remote one.    // yarn模式下,hdfs不支持加载到内存,所以采用"spark.yarn.dist.forceDownloadSchemes"方案(在添加到YARN分布式缓存之前，文件将被下载到本地磁盘的逗号分隔列表。用于YARN服务不支持Spark支持的方案的情况)    // 所以先把方案列表文件下载到本地,再通过相应方案加载资源到分布式内存中    // 在yarn-client模式中,上面的代码中已经把远程文件下载到了本地,只需要获取本地路径替换掉远程路径即可    if (clusterManager == YARN) {      // 加载方案列表      val forceDownloadSchemes = sparkConf.get(FORCE_DOWNLOAD_SCHEMES)      // 判断是否需要下载的方法      def shouldDownload(scheme: String): Boolean = {        forceDownloadSchemes.contains("*") || forceDownloadSchemes.contains(scheme) ||          Try {            FileSystem.getFileSystemClass(scheme, hadoopConf)          }.isFailure      }      // 下载资源的方法      def downloadResource(resource: String): String = {        val uri = Utils.resolveURI(resource)        uri.getScheme match {          case "local" | "file" => resource          case e if shouldDownload(e) =>            val file = new File(targetDir, new Path(uri).getName)            if (file.exists()) {              file.toURI.toString            } else {              downloadFile(resource, targetDir, sparkConf, hadoopConf, secMgr)            }          case _ => uri.toString        }      }      // 下载主要运行资源      args.primaryResource = Option(args.primaryResource).map {        downloadResource      }.orNull      // 下载文件      args.files = Option(args.files).map { files =>        Utils.stringToSeq(files).map(downloadResource).mkString(",")      }.orNull      args.pyFiles = Option(args.pyFiles).map { pyFiles =>        Utils.stringToSeq(pyFiles).map(downloadResource).mkString(",")      }.orNull      // 下载jars      args.jars = Option(args.jars).map { jars =>        Utils.stringToSeq(jars).map(downloadResource).mkString(",")      }.orNull      // 下载压缩文件      args.archives = Option(args.archives).map { archives =>        Utils.stringToSeq(archives).map(downloadResource).mkString(",")      }.orNull    }    // 如果我们正在运行python应用，请将主类设置为特定的python运行器    if (args.isPython && deployMode == CLIENT) {      if (args.primaryResource == PYSPARK_SHELL) {        args.mainClass = "org.apache.spark.api.python.PythonGatewayServer"      } else {        // If a python file is provided, add it to the child arguments and list of files to deploy.        // Usage: PythonAppRunner  [app arguments]        args.mainClass = "org.apache.spark.deploy.PythonRunner"        args.childArgs = ArrayBuffer(localPrimaryResource, localPyFiles) ++ args.childArgs        if (clusterManager != YARN) {          // The YARN backend distributes the primary file differently, so don't merge it.          args.files = mergeFileLists(args.files, args.primaryResource)        }      }      if (clusterManager != YARN) {        // The YARN backend handles python files differently, so don't merge the lists.        args.files = mergeFileLists(args.files, args.pyFiles)      }      if (localPyFiles != null) {        sparkConf.set("spark.submit.pyFiles", localPyFiles)      }    }    // 在R应用程序的yarn模式中，添加SparkR包存档和包含所有构建的R库的R包存档到存档中，以便它们可以随作业一起分发    if (args.isR && clusterManager == YARN) {      val sparkRPackagePath = RUtils.localSparkRPackagePath      if (sparkRPackagePath.isEmpty) {        printErrorAndExit("SPARK_HOME does not exist for R application in YARN mode.")      }      val sparkRPackageFile = new File(sparkRPackagePath.get, SPARKR_PACKAGE_ARCHIVE)      if (!sparkRPackageFile.exists()) {        printErrorAndExit(s"$SPARKR_PACKAGE_ARCHIVE does not exist for R application in YARN mode.")      }      val sparkRPackageURI = Utils.resolveURI(sparkRPackageFile.getAbsolutePath).toString      // Distribute the SparkR package.      // Assigns a symbol link name "sparkr" to the shipped package.      args.archives = mergeFileLists(args.archives, sparkRPackageURI + "#sparkr")      // Distribute the R package archive containing all the built R packages.      if (!RUtils.rPackages.isEmpty) {        val rPackageFile =          RPackageUtils.zipRLibraries(new File(RUtils.rPackages.get), R_PACKAGE_ARCHIVE)        if (!rPackageFile.exists()) {          printErrorAndExit("Failed to zip all the built R packages.")        }        val rPackageURI = Utils.resolveURI(rPackageFile.getAbsolutePath).toString        // Assigns a symbol link name "rpkg" to the shipped package.        args.archives = mergeFileLists(args.archives, rPackageURI + "#rpkg")      }    }    // TODO: Support distributing R packages with standalone cluster    if (args.isR && clusterManager == STANDALONE && !RUtils.rPackages.isEmpty) {      printErrorAndExit("Distributing R packages with standalone cluster is not supported.")    }    // TODO: Support distributing R packages with mesos cluster    if (args.isR && clusterManager == MESOS && !RUtils.rPackages.isEmpty) {      printErrorAndExit("Distributing R packages with mesos cluster is not supported.")    }    // 如果我们正在运行R应用，请将主类设置为特定的R运行器    if (args.isR && deployMode == CLIENT) {      if (args.primaryResource == SPARKR_SHELL) {        args.mainClass = "org.apache.spark.api.r.RBackend"      } else {        // If an R file is provided, add it to the child arguments and list of files to deploy.        // Usage: RRunner  [app arguments]        args.mainClass = "org.apache.spark.deploy.RRunner"        args.childArgs = ArrayBuffer(localPrimaryResource) ++ args.childArgs        args.files = mergeFileLists(args.files, args.primaryResource)      }    }    if (isYarnCluster && args.isR) {      // In yarn-cluster mode for an R app, add primary resource to files      // that can be distributed with the job      args.files = mergeFileLists(args.files, args.primaryResource)    }    // Special flag to avoid deprecation warnings at the client    sys.props("SPARK_SUBMIT") = "true"    //  为各种部署模式设置相应参数这里返回的是元组OptionAssigner类没有方法,只是设置了参数类型    val options = List[OptionAssigner](      // All cluster managers      OptionAssigner(args.master, ALL_CLUSTER_MGRS, ALL_DEPLOY_MODES, confKey = "spark.master"),      OptionAssigner(args.deployMode, ALL_CLUSTER_MGRS, ALL_DEPLOY_MODES,        confKey = "spark.submit.deployMode"),      OptionAssigner(args.name, ALL_CLUSTER_MGRS, ALL_DEPLOY_MODES, confKey = "spark.app.name"),      OptionAssigner(args.ivyRepoPath, ALL_CLUSTER_MGRS, CLIENT, confKey = "spark.jars.ivy"),      OptionAssigner(args.driverMemory, ALL_CLUSTER_MGRS, CLIENT,        confKey = "spark.driver.memory"),      OptionAssigner(args.driverExtraClassPath, ALL_CLUSTER_MGRS, ALL_DEPLOY_MODES,        confKey = "spark.driver.extraClassPath"),      OptionAssigner(args.driverExtraJavaOptions, ALL_CLUSTER_MGRS, ALL_DEPLOY_MODES,        confKey = "spark.driver.extraJavaOptions"),      OptionAssigner(args.driverExtraLibraryPath, ALL_CLUSTER_MGRS, ALL_DEPLOY_MODES,        confKey = "spark.driver.extraLibraryPath"),      // Propagate attributes for dependency resolution at the driver side      OptionAssigner(args.packages, STANDALONE | MESOS, CLUSTER, confKey = "spark.jars.packages"),      OptionAssigner(args.repositories, STANDALONE | MESOS, CLUSTER,        confKey = "spark.jars.repositories"),      OptionAssigner(args.ivyRepoPath, STANDALONE | MESOS, CLUSTER, confKey = "spark.jars.ivy"),      OptionAssigner(args.packagesExclusions, STANDALONE | MESOS,        CLUSTER, confKey = "spark.jars.excludes"),      // Yarn only      OptionAssigner(args.queue, YARN, ALL_DEPLOY_MODES, confKey = "spark.yarn.queue"),      OptionAssigner(args.numExecutors, YARN, ALL_DEPLOY_MODES,        confKey = "spark.executor.instances"),      OptionAssigner(args.pyFiles, YARN, ALL_DEPLOY_MODES, confKey = "spark.yarn.dist.pyFiles"),      OptionAssigner(args.jars, YARN, ALL_DEPLOY_MODES, confKey = "spark.yarn.dist.jars"),      OptionAssigner(args.files, YARN, ALL_DEPLOY_MODES, confKey = "spark.yarn.dist.files"),      OptionAssigner(args.archives, YARN, ALL_DEPLOY_MODES, confKey = "spark.yarn.dist.archives"),      OptionAssigner(args.principal, YARN, ALL_DEPLOY_MODES, confKey = "spark.yarn.principal"),      OptionAssigner(args.keytab, YARN, ALL_DEPLOY_MODES, confKey = "spark.yarn.keytab"),      // Other options      OptionAssigner(args.executorCores, STANDALONE | YARN | KUBERNETES, ALL_DEPLOY_MODES,        confKey = "spark.executor.cores"),      OptionAssigner(args.executorMemory, STANDALONE | MESOS | YARN | KUBERNETES, ALL_DEPLOY_MODES,        confKey = "spark.executor.memory"),      OptionAssigner(args.totalExecutorCores, STANDALONE | MESOS | KUBERNETES, ALL_DEPLOY_MODES,        confKey = "spark.cores.max"),      OptionAssigner(args.files, LOCAL | STANDALONE | MESOS | KUBERNETES, ALL_DEPLOY_MODES,        confKey = "spark.files"),      OptionAssigner(args.jars, LOCAL, CLIENT, confKey = "spark.jars"),      OptionAssigner(args.jars, STANDALONE | MESOS | KUBERNETES, ALL_DEPLOY_MODES,        confKey = "spark.jars"),      OptionAssigner(args.driverMemory, STANDALONE | MESOS | YARN | KUBERNETES, CLUSTER,        confKey = "spark.driver.memory"),      OptionAssigner(args.driverCores, STANDALONE | MESOS | YARN | KUBERNETES, CLUSTER,        confKey = "spark.driver.cores"),      OptionAssigner(args.supervise.toString, STANDALONE | MESOS, CLUSTER,        confKey = "spark.driver.supervise"),      OptionAssigner(args.ivyRepoPath, STANDALONE, CLUSTER, confKey = "spark.jars.ivy"),      // An internal option used only for spark-shell to add user jars to repl's classloader,      // previously it uses "spark.jars" or "spark.yarn.dist.jars" which now may be pointed to      // remote jars, so adding a new option to only specify local jars for spark-shell internally.      OptionAssigner(localJars, ALL_CLUSTER_MGRS, CLIENT, confKey = "spark.repl.local.jars")    )    // 在客户端模式下，直接启动应用程序主类    // 另外，将主应用程序jar和所有添加的jar(如果有)添加到classpath    if (deployMode == CLIENT) {      childMainClass = args.mainClass      if (localPrimaryResource != null && isUserJar(localPrimaryResource)) {        childClasspath += localPrimaryResource      }      if (localJars != null) {        childClasspath ++= localJars.split(",")      }    }    // 添加主应用程序jar和任何添加到类路径的jar，以yarn客户端需要这些jar。    // 这里假设primaryResource和user jar都是本地jar，否则它不会被添加到yarn客户端的类路径中。    if (isYarnCluster) {      if (isUserJar(args.primaryResource)) {        childClasspath += args.primaryResource      }      if (args.jars != null) {        childClasspath ++= args.jars.split(",")      }    }    if (deployMode == CLIENT) {      if (args.childArgs != null) {        childArgs ++= args.childArgs      }    }    // 将所有参数映射到我们选择的模式的命令行选项或系统属性    for (opt  x.split(",").toSeq).getOrElse(Seq.empty)      if (isUserJar(args.primaryResource)) {        jars = jars ++ Seq(args.primaryResource)      }      sparkConf.set("spark.jars", jars.mkString(","))    }    // 在standalone cluster模式下，使用REST客户端提交应用程序(Spark 1.3+)。所有Spark参数都将通过系统属性传递给客户端。    if (args.isStandaloneCluster) {      if (args.useRest) {        childMainClass = REST_CLUSTER_SUBMIT_CLASS        childArgs += (args.primaryResource, args.mainClass)      } else {        // In legacy standalone cluster mode, use Client as a wrapper around the user class        childMainClass = STANDALONE_CLUSTER_SUBMIT_CLASS        if (args.supervise) {          childArgs += "--supervise"        }        Option(args.driverMemory).foreach { m => childArgs += ("--memory", m) }        Option(args.driverCores).foreach { c => childArgs += ("--cores", c) }        childArgs += "launch"        childArgs += (args.master, args.primaryResource, args.mainClass)      }      if (args.childArgs != null) {        childArgs ++= args.childArgs      }    }    // 让YARN知道这是一个pyspark应用程序，因此它将分发所需的库。    if (clusterManager == YARN) {      if (args.isPython) {        sparkConf.set("spark.yarn.isPython", "true")      }    }    if (clusterManager == MESOS && UserGroupInformation.isSecurityEnabled) {      setRMPrincipal(sparkConf)    }    // 在yarn-cluster模式下，将yarn.Client用作用户类的包装器    if (isYarnCluster) {      childMainClass = YARN_CLUSTER_SUBMIT_CLASS      if (args.isPython) {        childArgs += ("--primary-py-file", args.primaryResource)        childArgs += ("--class", "org.apache.spark.deploy.PythonRunner")      } else if (args.isR) {        val mainFile = new Path(args.primaryResource).getName        childArgs += ("--primary-r-file", mainFile)        childArgs += ("--class", "org.apache.spark.deploy.RRunner")      } else {        if (args.primaryResource != SparkLauncher.NO_RESOURCE) {          childArgs += ("--jar", args.primaryResource)        }        childArgs += ("--class", args.mainClass)      }      if (args.childArgs != null) {        args.childArgs.foreach { arg => childArgs += ("--arg", arg) }      }    }    if (isMesosCluster) {      assert(args.useRest, "Mesos cluster mode is only supported through the REST submission API")      childMainClass = REST_CLUSTER_SUBMIT_CLASS      if (args.isPython) {        // Second argument is main class        childArgs += (args.primaryResource, "")        if (args.pyFiles != null) {          sparkConf.set("spark.submit.pyFiles", args.pyFiles)        }      } else if (args.isR) {        // Second argument is main class        childArgs += (args.primaryResource, "")      } else {        childArgs += (args.primaryResource, args.mainClass)      }      if (args.childArgs != null) {        childArgs ++= args.childArgs      }    }    if (isKubernetesCluster) {      childMainClass = KUBERNETES_CLUSTER_SUBMIT_CLASS      if (args.primaryResource != SparkLauncher.NO_RESOURCE) {        childArgs ++= Array("--primary-java-resource", args.primaryResource)      }      childArgs ++= Array("--main-class", args.mainClass)      if (args.childArgs != null) {        args.childArgs.foreach { arg =>          childArgs += ("--arg", arg)        }      }    }    // 加载通过--conf和默认属性文件指定的所有属性    for ((k, v)       // 如果存在，用解析的URI替换旧的URI      sparkConf.getOption(config).foreach { oldValue =>        sparkConf.set(config, Utils.resolveURIs(oldValue))      }    }    // 清理和格式化python文件的路径    // 如果默认配置中有设置spark.submit.pyFiles,name--py-files不用添加    sparkConf.getOption("spark.submit.pyFiles").foreach { pyFiles =>      val resolvedPyFiles = Utils.resolveURIs(pyFiles)      val formattedPyFiles = if (!isYarnCluster && !isMesosCluster) {        PythonRunner.formatPaths(resolvedPyFiles).mkString(",")      } else {        // 返回清理和格式化后的python文件路径        resolvedPyFiles      }      sparkConf.set("spark.submit.pyFiles", formattedPyFiles)    }    // 最终prepareSubmitEnvironment()返回的元组,对应了(mainclass args, jars classpath, sparkConf, mainclass path)    (childArgs, childClasspath, sparkConf, childMainClass)  }

2.4.2.4 doRunMain

    // 主要是调用runMain()启动相应环境的main()的方法    // 环境准备好以后,会先往下运行判断,这里是在等着调用    def doRunMain(): Unit = {      // 提交时可以指定--proxy-user,如果没有指定,则获取当前用户      if (args.proxyUser != null) {        val proxyUser = UserGroupInformation.createProxyUser(args.proxyUser,          UserGroupInformation.getCurrentUser())        try {          proxyUser.doAs(new PrivilegedExceptionAction[Unit]() {            override def run(): Unit = {              // 这里是真正的执行,runMain()              runMain(childArgs, childClasspath, sparkConf, childMainClass, args.verbose)            }          })        } catch {          case e: Exception =>            // Hadoop's AuthorizationException suppresses the exception's stack trace, which            // makes the message printed to the output by the JVM not very helpful. Instead,            // detect exceptions with empty stack traces here, and treat them differently.            if (e.getStackTrace().length == 0) {              // scalastyle:off println              printStream.println(s"ERROR: ${e.getClass().getName()}: ${e.getMessage()}")              // scalastyle:on println              exitFn(1)            } else {              throw e            }        }      } else {        // 没有指定用户时执行        runMain(childArgs, childClasspath, sparkConf, childMainClass, args.verbose)      }    }

2.4.2.5 runMain

/** * 使用提供的启动环境运行子类的main方法。 * 请注意，如果我们正在运行集群部署模式或python应用程序，则该主类将不是用户提供的主类。 * * 这里的参数有子类需要的参数,子类路径,sparkConf,子类main()路径,参数重复判断 */private def runMain(                     childArgs: Seq[String],                     childClasspath: Seq[String],                     sparkConf: SparkConf,                     childMainClass: String,                     verbose: Boolean): Unit = {  if (verbose) {    printStream.println(s"Main class:$childMainClass")    printStream.println(s"Arguments:${childArgs.mkString("")}")    printStream.println(s"Spark config:${Utils.redact(sparkConf.getAll.toMap).mkString("")}")    printStream.println(s"Classpath elements:${childClasspath.mkString("")}")    printStream.println("")  }  // 初始化类加载器  val loader = if (sparkConf.get(DRIVER_USER_CLASS_PATH_FIRST)) {    // 如果用户设置了class,通过ChildFirstURLClassLoader来加载    new ChildFirstURLClassLoader(new Array[URL](0), Thread.currentThread.getContextClassLoader)  } else {    // 如果用户没有设置,通过MutableURLClassLoader来加载    new MutableURLClassLoader(new Array[URL](0), Thread.currentThread.getContextClassLoader)  }  // 设置由上面自定义的类加载器来加载class到JVM  Thread.currentThread.setContextClassLoader(loader)  // 从Classpath中添加jars  for (jar       e.printStackTrace(printStream)      if (childMainClass.contains("thriftserver")) {        printStream.println(s"Failed to load main class $childMainClass.")        printStream.println("You need to build Spark with -Phive and -Phive-thriftserver.")      }      System.exit(CLASS_NOT_FOUND_EXIT_STATUS)    case e: NoClassDefFoundError =>      e.printStackTrace(printStream)      if (e.getMessage.contains("org/apache/hadoop/hive")) {        printStream.println(s"Failed to load hive class.")        printStream.println("You need to build Spark with -Phive and -Phive-thriftserver.")      }      System.exit(CLASS_NOT_FOUND_EXIT_STATUS)  }  /**   * 通过classOf[]构建从属于mainClass的SparkApplication对象   * 然后通过mainclass实例化了SparkApplication   * SparkApplication是一个抽象类,这里主要是实现它的start()   */  val app: SparkApplication = if (classOf[SparkApplication].isAssignableFrom(mainClass)) {    mainClass.newInstance().asInstanceOf[SparkApplication]  } else {    // SPARK-4170    if (classOf[scala.App].isAssignableFrom(mainClass)) {      printWarning("Subclasses of scala.App may not work correctly. Use a main() method instead.")    }    // 如果mainclass无法实例化SparkApplication,则使用替代构建子类JavaMainApplication实例    new JavaMainApplication(mainClass)  }  @tailrec  def findCause(t: Throwable): Throwable = t match {    case e: UndeclaredThrowableException =>      if (e.getCause() != null) findCause(e.getCause()) else e    case e: InvocationTargetException =>      if (e.getCause() != null) findCause(e.getCause()) else e    case e: Throwable =>      e  }  try {    // 启动实例    app.start(childArgs.toArray, sparkConf)  } catch {    case t: Throwable =>      findCause(t) match {        case SparkUserAppException(exitCode) =>          System.exit(exitCode)        case t: Throwable =>          throw t      }  }}

2.4.3 SparkApplication

package org.apache.spark.deployimport java.lang.reflect.Modifierimport org.apache.spark.SparkConf/** * 这是spark任务的入口抽象类,需要实现它的无参构造 */private[spark] trait SparkApplication {  def start(args: Array[String], conf: SparkConf): Unit}/** * 用main方法包装标准java类的SparkApplication实现 * * 用main方法包装标准java类的SparkApplication实现配置是通过系统配置文件传递,在同一个JVM中加载太多配置会可能导致配置溢出 */private[deploy] class JavaMainApplication(klass: Class[_]) extends SparkApplication {  override def start(args: Array[String], conf: SparkConf): Unit = {    val mainMethod = klass.getMethod("main", new Array[String](0).getClass)    if (!Modifier.isStatic(mainMethod.getModifiers)) {      throw new IllegalStateException("The main method in the given main class must be static")    }    val sysProps = conf.getAll.toMap    sysProps.foreach { case (k, v) =>      sys.props(k) = v    }    mainMethod.invoke(null, args)  }}

如果是在本地模式，到SparkApplication这个类这里已经运行结束。

但是如果是yarn cluster模式，它创建的实例是不同的，也就是start()启动的类其实是YarnClusterApplication，同样继承了SparkApplication，在后续的文章中回继续跟进。

3. 源码地址

https://github.com/perkinls/spark-2.3.3

4. 参考文献

《Spark内核设计艺术》关注公众号Data Porter 回复: Spark内核设计艺术免费领取

https://github.com/apache/spark

https://github.com/CrestOfWave/Spark-2.3.1

https://blog.csdn.net/do_yourself_go_on/article/details/75005204

https://blog.csdn.net/lingeio/article/details/96900714

欢迎公众号：Data Porter 免费获取数据结构、Java、Scala、Python、大数据、区块链、机器学习等学习资料。好手不敌双拳，双拳不如四手！希望认识更多的朋友一起成长、共同进步！