CDH集群spark-shell执行过程分析

目的

刚入门spark，安装的是CDH的版本，版本号spark-core_2.11-2.4.0-cdh6.2.1，部署了cdh客户端（非集群节点），本文主要以spark-shell为例子，对在cdh客户端上提交spark作业原理进行简单分析，加深理解

spark-shell执行

启动spark-shell后，可以发下yarn集群上启动了一个作业，实际上，cdh-spark默认提交作业模式为yarn-client模式，即在本地运行Driver，作业在yarn集群上执行

spark-shell启动过程分析

查看spark-shell路径及内容，$LIB_DIR值为/opt/cloudera/parcels/CDH/lib，所以执行的是/opt/cloudera/parcels/CDH/lib/spark/bin/spark-shell

继续查看/opt/cloudera/parcels/CDH/lib/spark/bin/spark-shell，脚本关键的内容如下：

#!/usr/bin/env bash
if [ -z "${SPARK_HOME}" ]; then
  source "$(dirname "$0")"/find-spark-home
fi
export _SPARK_CMD_USAGE="Usage: ./bin/spark-shell [options]"
SPARK_SUBMIT_OPTS="$SPARK_SUBMIT_OPTS -Dscala.usejavacp=true"
function main() {
    export SPARK_SUBMIT_OPTS
    "${SPARK_HOME}"/bin/spark-submit --class org.apache.spark.repl.Main --name "Spark shell" "$@"
  fi
}
main "$@"

上述脚本中首先判断是否存在SPARK_HOME变量，如果不存在的话就执行同一目录下的find-spark-home脚本，改脚本中如果存在SPARK_HOME存在，则直接返回。如果不返回，则查看当前目录下，是否有find_spark_home.py文件。如果存在find_spark_home.py文件，则调用python执行获取结果。如果不存在，则使用当前bin目录的上一级为SPARK_HOME，在本环境中SPARK_HOME被设置为/opt/cloudera/parcels/CDH/lib/spark，设置好SPARK_HOME之后，调用了spark-submit脚本。

查看spark-submit脚本，发现其调用的是${SPARK_HOME}"/bin/spark-class org.apache.spark.deploy.SparkSubmit，继续查看spark-class脚本，主要内容如下：

if [ -z "${SPARK_HOME}" ]; then
  source "$(dirname "$0")"/find-spark-home
fi

. "${SPARK_HOME}"/bin/load-spark-env.sh

# Find the java binary
if [ -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
  fi
fi

# Find Spark jars.
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"
fi

if [ ! -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 1
else
  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


build_command() {
  "$RUNNER" -Xmx128m -cp "$LAUNCH_CLASSPATH" org.apache.spark.launcher.Main "$@"
  printf "%d" $?
}

# Turn off posix mode since it does not allow process substitution
set +o posix
CMD=()
while IFS= read -d '' -r ARG; do
  CMD+=("$ARG")
done < <(build_command "$@")

COUNT=${#CMD[@]}
LAST=$((COUNT - 1))
LAUNCHER_EXIT_CODE=${CMD[$LAST]}

spark-class中，首先设置了spark-home，然后执行load-spark-env.sh，并添加${SPARK_HOME}/jars目录下的spark依赖，最后执行的是org.apache.spark.launcher.Main类，继续查看load-spark-env.sh
，改脚本主要是设置一些环境变量，关键内容如下：首先是设置spark_home，然后设置${SPARK_CONF_DIR}，并执行该目录下的spark-env.sh，SPARK_CONF_DIR默认为spark-home下的的conf目录，本环境为/opt/cloudera/parcels/CDH/lib/spark/conf

if [ -z "${SPARK_HOME}" ]; then
  source "$(dirname "$0")"/find-spark-home
fi

# Save SPARK_HOME in case the user's spark-env.sh overwrites it.
ORIGINAL_SPARK_HOME="$SPARK_HOME"

if [ -z "$SPARK_ENV_LOADED" ]; then
  export SPARK_ENV_LOADED=1

  export SPARK_CONF_DIR="${SPARK_CONF_DIR:-"${SPARK_HOME}"/conf}"

  if [ -f "${SPARK_CONF_DIR}/spark-env.sh" ]; then
    # Promote all variable declarations to environment (exported) variables
    set -a
    . "${SPARK_CONF_DIR}/spark-env.sh"
    set +a
  fi
fi

继续查看spark-env.sh内容，改脚本中直接设置了spark_home和hadoop_home目录，另外比较重要的是HADOOP_CONF_DIR和HIVE_CONF_DIR，如果没有设置的话，默认为cdh中提供配置文件，否则为用户设置的值，我们的环境bashrc中都设置了这两个变量，因此运行spark-shell时，会知道yarn集群的信息，建议使用spark-sql以及yarn模式运行作业是设置这两个变量

#!/usr/bin/env bash


SELF="$(cd $(dirname $BASH_SOURCE) && pwd)"
if [ -z "$SPARK_CONF_DIR" ]; then
  export SPARK_CONF_DIR="$SELF"
fi

export SPARK_HOME=/opt/cloudera/parcels/CDH-6.2.1-1.cdh6.2.1.p0.1425774/lib/spark

SPARK_PYTHON_PATH=""
if [ -n "$SPARK_PYTHON_PATH" ]; then
  export PYTHONPATH="$PYTHONPATH:$SPARK_PYTHON_PATH"
fi

export HADOOP_HOME=/opt/cloudera/parcels/CDH-6.2.1-1.cdh6.2.1.p0.1425774/lib/hadoop
export HADOOP_COMMON_HOME="$HADOOP_HOME"

if [ -n "$HADOOP_HOME" ]; then
  LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${HADOOP_HOME}/lib/native
fi

SPARK_EXTRA_LIB_PATH=""
if [ -n "$SPARK_EXTRA_LIB_PATH" ]; then
  LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$SPARK_EXTRA_LIB_PATH
fi

export LD_LIBRARY_PATH

HADOOP_CONF_DIR=${HADOOP_CONF_DIR:-$SPARK_CONF_DIR/yarn-conf}
HIVE_CONF_DIR=${HIVE_CONF_DIR:-/etc/hive/conf}
if [ -d "$HIVE_CONF_DIR" ]; then
  HADOOP_CONF_DIR="$HADOOP_CONF_DIR:$HIVE_CONF_DIR"
fi
export HADOOP_CONF_DIR

PYLIB="$SPARK_HOME/python/lib"
if [ -f "$PYLIB/pyspark.zip" ]; then
  PYSPARK_ARCHIVES_PATH=
  for lib in "$PYLIB"/*.zip; do
    if [ -n "$PYSPARK_ARCHIVES_PATH" ]; then
      PYSPARK_ARCHIVES_PATH="$PYSPARK_ARCHIVES_PATH,local:$lib"
    else
      PYSPARK_ARCHIVES_PATH="local:$lib"
    fi
  done
  export PYSPARK_ARCHIVES_PATH
fi

if [ -f "$SELF/classpath.txt" ]; then
  export SPARK_DIST_CLASSPATH=$(paste -sd: "$SELF/classpath.txt")
fi