《Hadoop》之"踽踽独行"（八）Hadoop集群的启动脚本整理及守护线程源码

在上一章的伪分布式集群搭建中，我们使用start-dfs.sh脚本启动了集群环境，并且上传了一个文件到HDFS上，还使用了mapreduce程序对HDFS上的这个文件进行了单词统计。今天我们就来简单了解一下启动脚本的相关内容和HDFS的一些重要的默认配置属性。

一、启动脚本

hadoop的脚本/指令目录，就两个，一个是bin/，一个是sbin/。现在，就来看看几个比较重要的脚本/指令。

1、sbin/start-all.sh

# Start all hadoop daemons.  Run this on master node.
# 开启所有的hadoop守护进程，在主节点上运行

echo "This script is Deprecated. Instead use start-dfs.sh and start-yarn.sh"
#这个脚本已经被弃用，使用start-dfs.sh和start-yarn.sh替代

bin=`dirname "${BASH_SOURCE-$0}"`
bin=`cd "$bin"; pwd`

DEFAULT_LIBEXEC_DIR="$bin"/../libexec
HADOOP_LIBEXEC_DIR=${HADOOP_LIBEXEC_DIR:-$DEFAULT_LIBEXEC_DIR}
. $HADOOP_LIBEXEC_DIR/hadoop-config.sh
#运行libexe/hadoop-config.sh指令，加载配置文件

# start hdfs daemons if hdfs is present
if [ -f "${HADOOP_HDFS_HOME}"/sbin/start-dfs.sh ]; then
  "${HADOOP_HDFS_HOME}"/sbin/start-dfs.sh --config $HADOOP_CONF_DIR
   #运行 sbin/start-dfs.sh指令
fi

# start yarn daemons if yarn is present
if [ -f "${HADOOP_YARN_HOME}"/sbin/start-yarn.sh ]; then
  "${HADOOP_YARN_HOME}"/sbin/start-yarn.sh --config $HADOOP_CONF_DIR
  #运行 sbin/start-yarn.sh指令
fi

我们可以看到，这个脚本的内容不多，实际上被弃用了，只不过是在这个start-all.sh脚本中，先执行hadoop-config.sh指令加载hadoop的一些环境变量，然后再分别执行start-dfs.sh脚本和start-yarn.sh脚本。

从此可以看出，我们也可以直接执行start-dfs.sh脚本来启动hadoop集群，无需执行start-all.sh脚本而已。（如果配置了yarn,再执行start-yarn.sh脚本）。

2、libexec/hadoop-config.sh

this="${BASH_SOURCE-$0}"
common_bin=$(cd -P -- "$(dirname -- "$this")" && pwd -P)
script="$(basename -- "$this")"
this="$common_bin/$script"

[ -f "$common_bin/hadoop-layout.sh" ] && . "$common_bin/hadoop-layout.sh"

HADOOP_COMMON_DIR=${HADOOP_COMMON_DIR:-"share/hadoop/common"}
HADOOP_COMMON_LIB_JARS_DIR=${HADOOP_COMMON_LIB_JARS_DIR:-"share/hadoop/common/lib"}
HADOOP_COMMON_LIB_NATIVE_DIR=${HADOOP_COMMON_LIB_NATIVE_DIR:-"lib/native"}
HDFS_DIR=${HDFS_DIR:-"share/hadoop/hdfs"}
HDFS_LIB_JARS_DIR=${HDFS_LIB_JARS_DIR:-"share/hadoop/hdfs/lib"}
YARN_DIR=${YARN_DIR:-"share/hadoop/yarn"}
YARN_LIB_JARS_DIR=${YARN_LIB_JARS_DIR:-"share/hadoop/yarn/lib"}
MAPRED_DIR=${MAPRED_DIR:-"share/hadoop/mapreduce"}
MAPRED_LIB_JARS_DIR=${MAPRED_LIB_JARS_DIR:-"share/hadoop/mapreduce/lib"}

# the root of the Hadoop installation
# See HADOOP-6255 for directory structure layout
HADOOP_DEFAULT_PREFIX=$(cd -P -- "$common_bin"/.. && pwd -P)
HADOOP_PREFIX=${HADOOP_PREFIX:-$HADOOP_DEFAULT_PREFIX}
export HADOOP_PREFIX
............................
...........省略细节，看重点..............
....................................
#  调用 hadoop-env.sh加载其他环境变量
if [ -f "${HADOOP_CONF_DIR}/hadoop-env.sh" ]; then
  . "${HADOOP_CONF_DIR}/hadoop-env.sh"
fi

这个脚本的作用，其实就是配置了一些hadoop集群的所需要的环境变量而已，内部还执行了hadoop-env.sh脚本，加载其他的比较重要的环境变量，如jdk等等

3、sbin/start-dfs.sh

# Start hadoop dfs daemons.          #开启HDFS的相关守护线程
# Optinally upgrade or rollback dfs state.   #可选升级或回滚DFS状态
# Run this on master node.           #在主节点上运行这个脚本

#这是start-dfs.sh的用法                              单独启动一个clusterId
usage="Usage: start-dfs.sh [-upgrade|-rollback] [other options such as -clusterId]"

bin=`dirname "${BASH_SOURCE-$0}"`
bin=`cd "$bin"; pwd`

DEFAULT_LIBEXEC_DIR="$bin"/../libexec
HADOOP_LIBEXEC_DIR=${HADOOP_LIBEXEC_DIR:-$DEFAULT_LIBEXEC_DIR}
#使用hdfs-config.sh加载环境变量
. $HADOOP_LIBEXEC_DIR/hdfs-config.sh

# get arguments
if [[ $# -ge 1 ]]; then
  startOpt="$1"
  shift
  case "$startOpt" in
    -upgrade)
      nameStartOpt="$startOpt"
    ;;
    -rollback)
      dataStartOpt="$startOpt"
    ;;
    *)
      echo $usage
      exit 1
    ;;
  esac
fi

#Add other possible options
nameStartOpt="$nameStartOpt $@"

#---------------------------------------------------------
# namenodes

NAMENODES=$($HADOOP_PREFIX/bin/hdfs getconf -namenodes)

echo "Starting namenodes on [$NAMENODES]"

#执行hadoop-daemons.sh  调用bin/hdfs指令  启动namenode守护线程
"$HADOOP_PREFIX/sbin/hadoop-daemons.sh" \
  --config "$HADOOP_CONF_DIR" \
  --hostnames "$NAMENODES" \
  --script "$bin/hdfs" start namenode $nameStartOpt

#---------------------------------------------------------
# datanodes (using default slaves file)

if [ -n "$HADOOP_SECURE_DN_USER" ]; then
  echo \
    "Attempting to start secure cluster, skipping datanodes. " \
    "Run start-secure-dns.sh as root to complete startup."
else
  #执行hadoop-daemons.sh  调用bin/hdfs指令 启动datanode守护线程
  "$HADOOP_PREFIX/sbin/hadoop-daemons.sh" \
    --config "$HADOOP_CONF_DIR" \
    --script "$bin/hdfs" start datanode $dataStartOpt
fi

#---------------------------------------------------------
# secondary namenodes (if any)

SECONDARY_NAMENODES=$($HADOOP_PREFIX/bin/hdfs getconf -secondarynamenodes 2>/dev/null)

if [ -n "$SECONDARY_NAMENODES" ]; then
  echo "Starting secondary namenodes [$SECONDARY_NAMENODES]"
  #执行hadoop-daemons.sh  调用bin/hdfs指令 启动secondarynamenode守护线程
  "$HADOOP_PREFIX/sbin/hadoop-daemons.sh" \
      --config "$HADOOP_CONF_DIR" \
      --hostnames "$SECONDARY_NAMENODES" \
      --script "$bin/hdfs" start secondarynamenode
fi
...................................
............省略细节.................
...................................

# eof

在start-dfs.sh脚本中，先执行hdfs-config.sh脚本加载环境变量，然后通过hadoop-daemons.sh脚本又调用bin/hdfs指令来分别开启namenode、datanode以及secondarynamenode等守护进程。

如此我们也能发现，其实直接执行hadoop-daemons.sh脚本，配合其用法，也应该可以启动HDFS等相关守护进程。

4、sbin/hadoop-daemons.sh

# 在所有的从节点上运行hadoop指令
# Run a Hadoop command on all slave hosts.

#hadoop-daemons.sh脚本的用法，                                            
usage="Usage: hadoop-daemons.sh [--config confdir] [--hosts hostlistfile] [start|stop] command args..."

# if no args specified, show usage
if [ $# -le 1 ]; then
  echo $usage
  exit 1
fi

bin=`dirname "${BASH_SOURCE-$0}"`
bin=`cd "$bin"; pwd`

DEFAULT_LIBEXEC_DIR="$bin"/../libexec
HADOOP_LIBEXEC_DIR=${HADOOP_LIBEXEC_DIR:-$DEFAULT_LIBEXEC_DIR}
#调用hadoop-config.sh加载环境比那里
. $HADOOP_LIBEXEC_DIR/hadoop-config.sh

#调用sbin/slaves.sh脚本  加载配置文件，然后使用hadoop-daemon.sh脚本读取配置文件
exec "$bin/slaves.sh" --config $HADOOP_CONF_DIR cd "$HADOOP_PREFIX" \; "$bin/hadoop-daemon.sh" --config $HADOOP_CONF_DIR "$@"

参考hadoop-daemons.sh的使用方法，不难发现直接使用hadoop-daemons.sh脚本，然后配合指令，就可以启动相关守护线程，如：

hadoop-daemons.sh  start  namenode                        #启动主节点
hadoop-daemons.sh  start  datanode                          #启动从节点
hadoop-daemons.sh  start  secondarynamenode       #启动第二主节点

在这个脚本中，我们可以看到内部执行了slaves.sh脚本读取环境变量，然后再调用了hadoop-daemon.sh脚本读取相关配置信息并执行了hadoop指令。

5、sbin/slaves.sh

# Run a shell command on all slave hosts.
#
# Environment Variables
#
#   HADOOP_SLAVES    File naming remote hosts.
#     Default is ${HADOOP_CONF_DIR}/slaves.
#   HADOOP_CONF_DIR  Alternate conf dir. Default is ${HADOOP_PREFIX}/conf.
#   HADOOP_SLAVE_SLEEP Seconds to sleep between spawning remote commands.
#   HADOOP_SSH_OPTS Options passed to ssh when running remote commands.
##

#  使用方法
usage="Usage: slaves.sh [--config confdir] command..."

# if no args specified, show usage
if [ $# -le 0 ]; then
  echo $usage
  exit 1
fi

bin=`dirname "${BASH_SOURCE-$0}"`
bin=`cd "$bin"; pwd`

DEFAULT_LIBEXEC_DIR="$bin"/../libexec
HADOOP_LIBEXEC_DIR=${HADOOP_LIBEXEC_DIR:-$DEFAULT_LIBEXEC_DIR}
. $HADOOP_LIBEXEC_DIR/hadoop-config.sh              #读取环境变量

if [ -f "${HADOOP_CONF_DIR}/hadoop-env.sh" ]; then
  . "${HADOOP_CONF_DIR}/hadoop-env.sh"              #读取环境变量
fi

# Where to start the script, see hadoop-config.sh
# (it set up the variables based on command line options)
if [ "$HADOOP_SLAVE_NAMES" != '' ] ; then
  SLAVE_NAMES=$HADOOP_SLAVE_NAMES
else
  SLAVE_FILE=${HADOOP_SLAVES:-${HADOOP_CONF_DIR}/slaves}
  SLAVE_NAMES=$(cat "$SLAVE_FILE" | sed  's/#.*$//;/^$/d')
fi

# start the daemons
for slave in $SLAVE_NAMES ; do
 ssh $HADOOP_SSH_OPTS $slave $"${@// /\\ }" \
   2>&1 | sed "s/^/$slave: /" &
 if [ "$HADOOP_SLAVE_SLEEP" != "" ]; then
   sleep $HADOOP_SLAVE_SLEEP
 fi
done

这个脚本也就是加载环境变量，然后通过ssh连接从节点。

6、sbin/hadoop-daemon.sh

#!/usr/bin/env bash


# Runs a Hadoop command as a daemon.     以守护进程的形式运行hadoop命令

.....................
.....................、

# 使用方法    command就是hadoop指令，下面有判读
usage="Usage: hadoop-daemon.sh [--config <conf-dir>] [--hosts hostlistfile] [--script script] (start|stop) <hadoop-command> <args...>"

.....................
.....................

#使用hadoop-config.sh加载环境变量
DEFAULT_LIBEXEC_DIR="$bin"/../libexec
HADOOP_LIBEXEC_DIR=${HADOOP_LIBEXEC_DIR:-$DEFAULT_LIBEXEC_DIR}
. $HADOOP_LIBEXEC_DIR/hadoop-config.sh

#使用hadoop-env.sh加载环境变量
if [ -f "${HADOOP_CONF_DIR}/hadoop-env.sh" ]; then
  . "${HADOOP_CONF_DIR}/hadoop-env.sh"
fi

.....................
.....................

case $startStop in

  (start)

    [ -w "$HADOOP_PID_DIR" ] ||  mkdir -p "$HADOOP_PID_DIR"

    if [ -f $pid ]; then
      if kill -0 `cat $pid` > /dev/null 2>&1; then
        echo $command running as process `cat $pid`.  Stop it first.
        exit 1
      fi
    fi

    if [ "$HADOOP_MASTER" != "" ]; then
      echo rsync from $HADOOP_MASTER
      rsync -a -e ssh --delete --exclude=.svn --exclude='logs/*' --exclude='contrib/hod/logs/*' $HADOOP_MASTER/ "$HADOOP_PREFIX"
    fi

    hadoop_rotate_log $log
    echo starting $command, logging to $log
    cd "$HADOOP_PREFIX"

#判断command是什么指令，然后调用bin/hdfs指令 读取配置文件，执行相关指令
    case $command in
      namenode|secondarynamenode|datanode|journalnode|dfs|dfsadmin|fsck|balancer|zkfc)
        if [ -z "$HADOOP_HDFS_HOME" ]; then
          hdfsScript="$HADOOP_PREFIX"/bin/hdfs
        else
          hdfsScript="$HADOOP_HDFS_HOME"/bin/hdfs
        fi
        nohup nice -n $HADOOP_NICENESS $hdfsScript --config $HADOOP_CONF_DIR $command "$@" > "$log" 2>&1 < /dev/null &
      ;;
      (*)
        nohup nice -n $HADOOP_NICENESS $hadoopScript --config $HADOOP_CONF_DIR $command "$@" > "$log" 2>&1 < /dev/null &
      ;;
    esac
 ........................
........................
esac

在hadoop-daemon.sh脚本中，同样读取了环境变量，然后依据传入的参数$@（上一个脚本中）来判断要启动的hadoop的守护线程($command)，最后调用bin/hdfs指令 读取配置信息 并启动hadoop的守护线程。

7、bin/hdfs

这是一个指令，而非shell脚本。我们可以发现，在启动hadoop集群时，不管使用什么脚本，最终都指向了bin/hdfs这个指令，那么这个指令里到底是什么呢，我们来看一下，就明白了。

bin=`which $0`
bin=`dirname ${bin}`
bin=`cd "$bin" > /dev/null; pwd`

DEFAULT_LIBEXEC_DIR="$bin"/../libexec

HADOOP_LIBEXEC_DIR=${HADOOP_LIBEXEC_DIR:-$DEFAULT_LIBEXEC_DIR}
. $HADOOP_LIBEXEC_DIR/hdfs-config.sh

#除了上面继续加载环境变化外，这个函数其实就是提示我们在使用什么
#比如namenode -format  是格式化DFS filesystem
#再比如 namenode    说的是运行一个DFS namenode
# 我们往下看
function print_usage(){
  echo "Usage: hdfs [--config confdir] [--loglevel loglevel] COMMAND"
  echo "       where COMMAND is one of:"
  echo "  dfs                  run a filesystem command on the file systems supported in Hadoop."
  echo "  classpath            prints the classpath"
  echo "  namenode -format     format the DFS filesystem"
  echo "  secondarynamenode    run the DFS secondary namenode"
  echo "  namenode             run the DFS namenode"
  echo "  journalnode          run the DFS journalnode"
  echo "  zkfc                 run the ZK Failover Controller daemon"
  echo "  datanode             run a DFS datanode"
  echo "  dfsadmin             run a DFS admin client"
  echo "  haadmin              run a DFS HA admin client"
  echo "  fsck                 run a DFS filesystem checking utility"
  echo "  balancer             run a cluster balancing utility"
  echo "  jmxget               get JMX exported values from NameNode or DataNode."
  echo "  mover                run a utility to move block replicas across"
  echo "                       storage types"
  echo "  oiv                  apply the offline fsimage viewer to an fsimage"
  echo "  oiv_legacy           apply the offline fsimage viewer to an legacy fsimage"
  echo "  oev                  apply the offline edits viewer to an edits file"
  echo "  fetchdt              fetch a delegation token from the NameNode"
  echo "  getconf              get config values from configuration"
  echo "  groups               get the groups which users belong to"
  echo "  snapshotDiff         diff two snapshots of a directory or diff the"
  echo "                       current directory contents with a snapshot"
  echo "  lsSnapshottableDir   list all snapshottable dirs owned by the current user"
  echo "						Use -help to see options"
  echo "  portmap              run a portmap service"
  echo "  nfs3                 run an NFS version 3 gateway"
  echo "  cacheadmin           configure the HDFS cache"
  echo "  crypto               configure HDFS encryption zones"
  echo "  storagepolicies      list/get/set block storage policies"
  echo "  version              print the version"
  echo ""
  echo "Most commands print help when invoked w/o parameters."
  # There are also debug commands, but they don't show up in this listing.
}

if [ $# = 0 ]; then
  print_usage
  exit
fi

COMMAND=$1
shift

case $COMMAND in
  # usage flags
  --help|-help|-h)
    print_usage
    exit
    ;;
esac

# Determine if we're starting a secure datanode, and if so, redefine appropriate variables
if [ "$COMMAND" == "datanode" ] && [ "$EUID" -eq 0 ] && [ -n "$HADOOP_SECURE_DN_USER" ]; then
  if [ -n "$JSVC_HOME" ]; then
    if [ -n "$HADOOP_SECURE_DN_PID_DIR" ]; then
      HADOOP_PID_DIR=$HADOOP_SECURE_DN_PID_DIR
    fi
  
    if [ -n "$HADOOP_SECURE_DN_LOG_DIR" ]; then
      HADOOP_LOG_DIR=$HADOOP_SECURE_DN_LOG_DIR
      HADOOP_OPTS="$HADOOP_OPTS -Dhadoop.log.dir=$HADOOP_LOG_DIR"
    fi
   
    HADOOP_IDENT_STRING=$HADOOP_SECURE_DN_USER
    HADOOP_OPTS="$HADOOP_OPTS -Dhadoop.id.str=$HADOOP_IDENT_STRING"
    starting_secure_dn="true"
  else
    echo "It looks like you're trying to start a secure DN, but \$JSVC_HOME"\
      "isn't set. Falling back to starting insecure DN."
  fi
fi

# Determine if we're starting a privileged NFS daemon, and if so, redefine appropriate variables
if [ "$COMMAND" == "nfs3" ] && [ "$EUID" -eq 0 ] && [ -n "$HADOOP_PRIVILEGED_NFS_USER" ]; then
  if [ -n "$JSVC_HOME" ]; then
    if [ -n "$HADOOP_PRIVILEGED_NFS_PID_DIR" ]; then
      HADOOP_PID_DIR=$HADOOP_PRIVILEGED_NFS_PID_DIR
    fi

    if [ -n "$HADOOP_PRIVILEGED_NFS_LOG_DIR" ]; then
      HADOOP_LOG_DIR=$HADOOP_PRIVILEGED_NFS_LOG_DIR
      HADOOP_OPTS="$HADOOP_OPTS -Dhadoop.log.dir=$HADOOP_LOG_DIR"
    fi
   
    HADOOP_IDENT_STRING=$HADOOP_PRIVILEGED_NFS_USER
    HADOOP_OPTS="$HADOOP_OPTS -Dhadoop.id.str=$HADOOP_IDENT_STRING"
    starting_privileged_nfs="true"
  else
    echo "It looks like you're trying to start a privileged NFS server, but"\
      "\$JSVC_HOME isn't set. Falling back to starting unprivileged NFS server."
  fi
fi

#   停停停，对就是这
#   我们可以看到，通过相应的hadoop指令，在加载相应的class文件
#   然后在jvm运行此程序。别忘记了，hadoop是用java语言开发的

if [ "$COMMAND" = "namenode" ] ; then
  CLASS='org.apache.hadoop.hdfs.server.namenode.NameNode'    #namenode守护线程对应的CLASS字节码
  HADOOP_OPTS="$HADOOP_OPTS $HADOOP_NAMENODE_OPTS"
elif [ "$COMMAND" = "zkfc" ] ; then
  CLASS='org.apache.hadoop.hdfs.tools.DFSZKFailoverController'
  HADOOP_OPTS="$HADOOP_OPTS $HADOOP_ZKFC_OPTS"
elif [ "$COMMAND" = "secondarynamenode" ] ; then
  CLASS='org.apache.hadoop.hdfs.server.namenode.SecondaryNameNode'    #SecondaryNameNode守护线程对应的CLASS字节码
  HADOOP_OPTS="$HADOOP_OPTS $HADOOP_SECONDARYNAMENODE_OPTS"
elif [ "$COMMAND" = "datanode" ] ; then
  CLASS='org.apache.hadoop.hdfs.server.datanode.DataNode'           #DataNode守护线程对应的CLASS字节码
  if [ "$starting_secure_dn" = "true" ]; then
    HADOOP_OPTS="$HADOOP_OPTS -jvm server $HADOOP_DATANODE_OPTS"
  else
    HADOOP_OPTS="$HADOOP_OPTS -server $HADOOP_DATANODE_OPTS"
  fi
elif [ "$COMMAND" = "journalnode" ] ; then
  CLASS='org.apache.hadoop.hdfs.qjournal.server.JournalNode'
  HADOOP_OPTS="$HADOOP_OPTS $HADOOP_JOURNALNODE_OPTS"

.......................................
...............省略很多..............
.......................................
# Check to see if we should start a secure datanode
if [ "$starting_secure_dn" = "true" ]; then
  if [ "$HADOOP_PID_DIR" = "" ]; then
    HADOOP_SECURE_DN_PID="/tmp/hadoop_secure_dn.pid"
  else
    HADOOP_SECURE_DN_PID="$HADOOP_PID_DIR/hadoop_secure_dn.pid"
  fi

  JSVC=$JSVC_HOME/jsvc
  if [ ! -f $JSVC ]; then
    echo "JSVC_HOME is not set correctly so jsvc cannot be found. jsvc is required to run secure datanodes. "
    echo "Please download and install jsvc from http://archive.apache.org/dist/commons/daemon/binaries/ "\
      "and set JSVC_HOME to the directory containing the jsvc binary."
    exit
  fi

  if [[ ! $JSVC_OUTFILE ]]; then
    JSVC_OUTFILE="$HADOOP_LOG_DIR/jsvc.out"
  fi

  if [[ ! $JSVC_ERRFILE ]]; then
    JSVC_ERRFILE="$HADOOP_LOG_DIR/jsvc.err"
  fi
   #运行 java字节码文件
  exec "$JSVC" \
           -Dproc_$COMMAND -outfile "$JSVC_OUTFILE" \
           -errfile "$JSVC_ERRFILE" \
           -pidfile "$HADOOP_SECURE_DN_PID" \
           -nodetach \
           -user "$HADOOP_SECURE_DN_USER" \
            -cp "$CLASSPATH" \
           $JAVA_HEAP_MAX $HADOOP_OPTS \
           org.apache.hadoop.hdfs.server.datanode.SecureDataNodeStarter "$@"
elif [ "$starting_privileged_nfs" = "true" ] ; then
  if [ "$HADOOP_PID_DIR" = "" ]; then
    HADOOP_PRIVILEGED_NFS_PID="/tmp/hadoop_privileged_nfs3.pid"
  else
    HADOOP_PRIVILEGED_NFS_PID="$HADOOP_PID_DIR/hadoop_privileged_nfs3.pid"
  fi

  JSVC=$JSVC_HOME/jsvc
  if [ ! -f $JSVC ]; then
    echo "JSVC_HOME is not set correctly so jsvc cannot be found. jsvc is required to run privileged NFS gateways. "
    echo "Please download and install jsvc from http://archive.apache.org/dist/commons/daemon/binaries/ "\
      "and set JSVC_HOME to the directory containing the jsvc binary."
    exit
  fi

  if [[ ! $JSVC_OUTFILE ]]; then
    JSVC_OUTFILE="$HADOOP_LOG_DIR/nfs3_jsvc.out"
  fi

  if [[ ! $JSVC_ERRFILE ]]; then
    JSVC_ERRFILE="$HADOOP_LOG_DIR/nfs3_jsvc.err"
  fi

   #运行 java字节码文件
  exec "$JSVC" \
           -Dproc_$COMMAND -outfile "$JSVC_OUTFILE" \
           -errfile "$JSVC_ERRFILE" \
           -pidfile "$HADOOP_PRIVILEGED_NFS_PID" \
           -nodetach \
           -user "$HADOOP_PRIVILEGED_NFS_USER" \
           -cp "$CLASSPATH" \
           $JAVA_HEAP_MAX $HADOOP_OPTS \
           org.apache.hadoop.hdfs.nfs.nfs3.PrivilegedNfsGatewayStarter "$@"
else

  #运行 java字节码文件
  # run it
  exec "$JAVA" -Dproc_$COMMAND $JAVA_HEAP_MAX $HADOOP_OPTS $CLASS "$@"
fi

看完懂了吗？在这个指令中，加载了各个守护线程对应的CLASS字节码文件，然后在JVM上来运行相应的守护线程。

hadoop的另一个指令bin/hadoop，内部也调用了bin/hdfs指令，感兴趣的话，可以自己看看，我就不展示出来了。至于跟yarn有关的脚本和指令也是相同的逻辑关系，我也不一一展示了。

使用图片重写整理了一下启动脚本的执行先后顺序：

使用文字再次整理一下：

#一个脚本启动所有线程
start-all.sh         #执行此脚本可以启动所有线程
    1. hadoop-config.sh
        a. hadoop-env.sh    
    2. start-dfs.sh  #执行此脚本可以启动HDFS相关线程
        a.hadoop-config.sh
        b.hadoop-daemons.sh   hdfs  namenode
          hadoop-daemons.sh   hdfs  datanode
          hadoop-daemons.sh   hdfs  secondarynamenode   
    3. start-yarn.sh  #执行此脚本可以启动YARN相关线程

#启动单个线程
#方法1：
hadoop-daemons.sh   --config   [start|stop]  command
    1. hadoop-config.sh
        a. hadoop-env.sh
    2. slaves.sh
        a. hadoop-config.sh
        b. hadoop-env.sh
    3. hadoop-daemon.sh --config  [start|stop]  command
        a.hdfs $command
#方法2：
hadoop-daemon.sh   --config   [start|stop]  command
    1. hadoop-config.sh
        a. hadoop-env.sh
    2. hdfs $command

二、底层源码查看

我们通过捋顺启动脚本发现，启动namenode对应的字节码文件是：org.apache.hadoop.hdfs.server.namenode.NameNode。启动datanode对应的字节码文件是：org.apache.hadoop.hdfs.server.datanode.DataNode。而启动secondarynamenode对应的字节码文件是：org.apache.hadoop.hdfs.server.namenode.SecondaryNameNode。

这些源码所在的har包：hadoop-hdfs-2.7.3-sources.jar

1、namenode的源码

package org.apache.hadoop.hdfs.server.namenode;

.......................
import org.apache.hadoop.hdfs.HdfsConfiguration;
..........................
@InterfaceAudience.Private
public class NameNode implements NameNodeStatusMXBean {
  static{   //静态块
    HdfsConfiguration.init();     //调用HdfsConfiguration的init方法，进行读取配置文件
  }
  ...................
  public static void main(String argv[]) throws Exception {
    if (DFSUtil.parseHelpArgument(argv, NameNode.USAGE, System.out, true)) {
      System.exit(0);
    }

    try {
      StringUtils.startupShutdownMessage(NameNode.class, argv, LOG);
      NameNode namenode = createNameNode(argv, null);  //创建namenode
      if (namenode != null) {
        namenode.join();   //启动namenode线程
      }
    } catch (Throwable e) {
      LOG.error("Failed to start namenode.", e);
      terminate(1, e);
    }
  }
...........
}

看一下HdfsConfiguration类

package org.apache.hadoop.hdfs;

/**
 * Adds deprecated keys into the configuration.
 */
@InterfaceAudience.Private
public class HdfsConfiguration extends Configuration {
  static {       //静态块
    addDeprecatedKeys();
    // adds the default resources
    Configuration.addDefaultResource("hdfs-default.xml");       //取默认配置文件
    Configuration.addDefaultResource("hdfs-site.xml");          //读取个人设置文件

  }
  
  public static void init() {}

  private static void addDeprecatedKeys() {}

  public static void main(String[] args) {
    init();
    Configuration.dumpDeprecatedKeys();
  }
}

2、datanode源码

package org.apache.hadoop.hdfs.server.datanode;
..............
import org.apache.hadoop.hdfs.HdfsConfiguration;
..............
@InterfaceAudience.Private
public class DataNode extends ReconfigurableBase
    implements InterDatanodeProtocol, ClientDatanodeProtocol,
        TraceAdminProtocol, DataNodeMXBean {
  public static final Log LOG = LogFactory.getLog(DataNode.class);
  
  static{
    HdfsConfiguration.init();    //同样在静态块中调用了HdfsConfiguration类，用于加载配置文件
  }
}

---------------------------------------如有疑问，敬请留言------------------------------------------