HDFS-2.7.0系列2: hadoop namenode -format腳本解析

咱們都知道，在Hadoop 2.7.0中，系統剛開始執行時，須要執行一個命令就是

hadoop namenode -format

下面咱們就來逐行分析下，這個命令對應的腳本的真正內涵！

---

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

打印出來的bin值爲

/root/hadoop-2.7.0-bin/bin

---

DEFAULT_LIBEXEC_DIR="$bin"/../libexec

則致使DEFAULT_LIBEXEC_DIR的值爲

 /root/hadoop-2.7.0-bin/bin/../libexec

---

cygwin=false
case "$(uname)" in
CYGWIN*) cygwin=true;;
esac

這個一看就知道跟cywin相關，也就是若是你在windows下執行這個腳本纔有意義，個人是linux,無視這段代碼！

---

HADOOP_LIBEXEC_DIR=${HADOOP_LIBEXEC_DIR:-$DEFAULT_LIBEXEC_DIR}

打印其值爲

/root/hadoop-2.7.0-bin/bin/../libexec

---

接下來執行一個腳本

. $HADOOP_LIBEXEC_DIR/hadoop-config.sh

打印的值是/root/hadoop-2.7.0-bin/bin/../libexec/hadoop-config.sh

實際上也就是

/root/hadoop-2.7.0-bin/libexec/hadoop-config.sh

因此先轉向這個腳本看看

|

|

|

---解析/root/hadoop-2.7.0-bin/libexec/hadoop-config.sh 

剛開始是

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

 打印出全部的值爲

this=/root/hadoop-2.7.0-bin/libexec/hadoop-config.sh
common_bin=/root/hadoop-2.7.0-bin/libexec
script=hadoop-config.sh

---

而後執行一段腳本

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

謝天謝地，這裏並無這個腳本，因此不須要執行

---

接下來是一大波賦值語句

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"}

打印值以下：

HADOOP_COMMON_DIR=share/hadoop/common
HADOOP_COMMON_LIB_JARS_DIR=share/hadoop/common/lib
HADOOP_COMMON_LIB_NATIVE_DIR=lib/native
HDFS_DIR=share/hadoop/hdfs
HDFS_LIB_JARS_DIR=share/hadoop/hdfs/lib
YARN_DIR=share/hadoop/yarn
YARN_LIB_JARS_DIR=share/hadoop/yarn/lib
MAPRED_DIR=share/hadoop/mapreduce
MAPRED_LIB_JARS_DIR=share/hadoop/mapreduce/lib

---

接下來是一小波賦值語句

HADOOP_DEFAULT_PREFIX=$(cd -P -- "$common_bin"/.. && pwd -P)
HADOOP_PREFIX=${HADOOP_PREFIX:-$HADOOP_DEFAULT_PREFIX}
export HADOOP_PREFIX

打印值以下：

HADOOP_DEFAULT_PREFIX=/root/hadoop-2.7.0-bin
HADOOP_PREFIX=/root/hadoop-2.7.0-bin

---

接下來是一段判斷語句

#check to see if the conf dir is given as an optional argument
if [ $# -gt 1 ]
then
    if [ "--config" = "$1" ]
   then
       shift
       confdir=$1
       if [ ! -d "$confdir" ]; then
                echo "Error: Cannot find configuration directory: $confdir"
                exit 1
             fi
       shift
       HADOOP_CONF_DIR=$confdir
    fi
fi

 執行結果以下：

本次命令下沒有執行

---

if [ $# -gt 1 ]
then
  if [ "--loglevel" = "$1" ]
  then
    shift
    HADOOP_LOGLEVEL=$1
    shift
  fi
fi
HADOOP_LOGLEVEL="${HADOOP_LOGLEVEL:-INFO}"

執行結果爲

 HADOOP_LOGLEVEL=INFO

---

# Allow alternate conf dir location.
if [ -e "${HADOOP_PREFIX}/conf/hadoop-env.sh" ]; then
  DEFAULT_CONF_DIR="conf"
else
  DEFAULT_CONF_DIR="etc/hadoop"
fi
export HADOOP_CONF_DIR="${HADOOP_CONF_DIR:-$HADOOP_PREFIX/$DEFAULT_CONF_DIR}"

打印爲

HADOOP_CONF_DIR=/root/hadoop-2.7.0-bin/etc/hadoop

---

接下來是判斷腳本

# User can specify hostnames or a file where the hostnames are (not both)
if [[ ( "$HADOOP_SLAVES" != '' ) && ( "$HADOOP_SLAVE_NAMES" != '' ) ]] ; then
  echo \
    "Error: Please specify one variable HADOOP_SLAVES or " \
    "HADOOP_SLAVE_NAME and not both."
  exit 1
fi

結果爲

經過

---

# Process command line options that specify hosts or file with host
# list
if [ $# -gt 1 ]
then
    if [ "--hosts" = "$1" ]
    then
        shift
        export HADOOP_SLAVES="${HADOOP_CONF_DIR}/$1"
        shift
    elif [ "--hostnames" = "$1" ]
    then
        shift
        export HADOOP_SLAVE_NAMES=$1
        shift
    fi
fi

執行結果爲

很抱歉，任何一個分支都沒有被執行

---

# User can specify hostnames or a file where the hostnames are (not both)
# (same check as above but now we know it's command line options that cause
# the problem)
if [[ ( "$HADOOP_SLAVES" != '' ) && ( "$HADOOP_SLAVE_NAMES" != '' ) ]] ; then
  echo \
    "Error: Please specify one of --hosts or --hostnames options and not both."
  exit 1
fi

顯然執行結果爲不執行

---

接下來又執行另外一個腳本

if [ -f "${HADOOP_CONF_DIR}/hadoop-env.sh" ]; then
  . "${HADOOP_CONF_DIR}/hadoop-env.sh"
fi

路徑爲：

/root/hadoop-2.7.0-bin/etc/hadoop/hadoop-env.sh

下面來執行這個腳本

|

|

|

---/root/hadoop-2.7.0-bin/etc/hadoop/hadoop-env.sh

這個腳本，所有是export,就不細細分析了。

讓咱們再回到hadoop-config.sh

---

接下來是

# check if net.ipv6.bindv6only is set to 1
bindv6only=$(/sbin/sysctl -n net.ipv6.bindv6only 2> /dev/null)
if [ -n "$bindv6only" ] && [ "$bindv6only" -eq "1" ] && [ "$HADOOP_ALLOW_IPV6" != "yes" ]
then
  echo "Error: \"net.ipv6.bindv6only\" is set to 1 - Java networking could be broken"
  echo "For more info: http://wiki.apache.org/hadoop/HadoopIPv6"
  exit 1
fi

顯然不執行

---

接下來是

# Newer versions of glibc use an arena memory allocator that causes virtual
# memory usage to explode. This interacts badly with the many threads that
# we use in Hadoop. Tune the variable down to prevent vmem explosion.
export MALLOC_ARENA_MAX=${MALLOC_ARENA_MAX:-4}

打印

MALLOC_ARENA_MAX=4

---

接下來是JAVA_HOME

# Attempt to set JAVA_HOME if it is not set
if [[ -z $JAVA_HOME ]]; then
  # On OSX use java_home (or /Library for older versions)
  if [ "Darwin" == "$(uname -s)" ]; then
    if [ -x /usr/libexec/java_home ]; then
      export JAVA_HOME=($(/usr/libexec/java_home))
    else
      export JAVA_HOME=(/Library/Java/Home)
    fi
  fi
  # Bail if we did not detect it
  if [[ -z $JAVA_HOME ]]; then
    echo "Error: JAVA_HOME is not set and could not be found." 1>&2
    exit 1
  fi
fi

打印

JAVA_HOME=/usr/java/jdk1.8.0_45

---

JAVA=$JAVA_HOME/bin/java
# some Java parameters
JAVA_HEAP_MAX=-Xmx1000m 
# check envvars which might override default args
if [ "$HADOOP_HEAPSIZE" != "" ]; then
  #echo "run with heapsize $HADOOP_HEAPSIZE"
  JAVA_HEAP_MAX="-Xmx""$HADOOP_HEAPSIZE""m"
  #echo $JAVA_HEAP_MAX
fi

打印值爲

JAVA_HEAP_MAX=-Xmx1000m

---

# CLASSPATH initially contains $HADOOP_CONF_DIR
CLASSPATH="${HADOOP_CONF_DIR}"

打印

 CLASSPATH=/root/hadoop-2.7.0-bin/etc/hadoop

---

# so that filenames w/ spaces are handled correctly in loops below
IFS=
if [ "$HADOOP_COMMON_HOME" = "" ]; then
  if [ -d "${HADOOP_PREFIX}/$HADOOP_COMMON_DIR" ]; then
    export HADOOP_COMMON_HOME=$HADOOP_PREFIX
  fi
fi

打印

HADOOP_COMMON_HOME=/root/hadoop-2.7.0-bin

---

# for releases, add core hadoop jar & webapps to CLASSPATH
if [ -d "$HADOOP_COMMON_HOME/$HADOOP_COMMON_DIR/webapps" ]; then
  CLASSPATH=${CLASSPATH}:$HADOOP_COMMON_HOME/$HADOOP_COMMON_DIR
fi
if [ -d "$HADOOP_COMMON_HOME/$HADOOP_COMMON_LIB_JARS_DIR" ]; then
  CLASSPATH=${CLASSPATH}:$HADOOP_COMMON_HOME/$HADOOP_COMMON_LIB_JARS_DIR'/*'
fi
CLASSPATH=${CLASSPATH}:$HADOOP_COMMON_HOME/$HADOOP_COMMON_DIR'/*'

打印值

CLASSPATH=/root/hadoop-2.7.0-bin/etc/hadoop:/root/hadoop-2.7.0-bin/share/hadoop/common/lib/*:/root/hadoop-2.7.0-bin/share/hadoop/common/*

---

# default log directory & file
if [ "$HADOOP_LOG_DIR" = "" ]; then
  HADOOP_LOG_DIR="$HADOOP_PREFIX/logs"
fi
if [ "$HADOOP_LOGFILE" = "" ]; then
  HADOOP_LOGFILE='hadoop.log'
fi
# default policy file for service-level authorization
if [ "$HADOOP_POLICYFILE" = "" ]; then
  HADOOP_POLICYFILE="hadoop-policy.xml"
fi
# restore ordinary behaviour
unset IFS

打印

HADOOP_LOG_DIR=/root/hadoop-2.7.0-bin/logs
HADOOP_LOGFILE=hadoop.log
HADOOP_POLICYFILE=hadoop-policy.xml

---

# setup 'java.library.path' for native-hadoop code if necessary
if [ -d "${HADOOP_PREFIX}/build/native" -o -d "${HADOOP_PREFIX}/$HADOOP_COMMON_LIB_NATIVE_DIR" ]; then
    
  if [ -d "${HADOOP_PREFIX}/$HADOOP_COMMON_LIB_NATIVE_DIR" ]; then
    if [ "x$JAVA_LIBRARY_PATH" != "x" ]; then
      JAVA_LIBRARY_PATH=${JAVA_LIBRARY_PATH}:${HADOOP_PREFIX}/$HADOOP_COMMON_LIB_NATIVE_DIR
    else
      JAVA_LIBRARY_PATH=${HADOOP_PREFIX}/$HADOOP_COMMON_LIB_NATIVE_DIR
    fi
  fi
fi

打印

 JAVA_LIBRARY_PATH=/root/hadoop-2.7.0-bin/lib/native

---

# setup a default TOOL_PATH
TOOL_PATH="${TOOL_PATH:-$HADOOP_PREFIX/share/hadoop/tools/lib/*}"
HADOOP_OPTS="$HADOOP_OPTS -Dhadoop.log.dir=$HADOOP_LOG_DIR"
HADOOP_OPTS="$HADOOP_OPTS -Dhadoop.log.file=$HADOOP_LOGFILE"
HADOOP_HOME=$HADOOP_PREFIX
if $cygwin; then
  HADOOP_HOME=$(cygpath -w "$HADOOP_HOME" 2>/dev/null)
fi
export HADOOP_HOME
HADOOP_OPTS="$HADOOP_OPTS -Dhadoop.home.dir=$HADOOP_HOME"
HADOOP_OPTS="$HADOOP_OPTS -Dhadoop.id.str=$HADOOP_IDENT_STRING"
HADOOP_OPTS="$HADOOP_OPTS -Dhadoop.root.logger=${HADOOP_ROOT_LOGGER:-${HADOOP_LOGLEVEL},console}"
if [ "x$JAVA_LIBRARY_PATH" != "x" ]; then
  if $cygwin; then
    JAVA_LIBRARY_PATH=$(cygpath -w "$JAVA_LIBRARY_PATH" 2>/dev/null)
  fi
  HADOOP_OPTS="$HADOOP_OPTS -Djava.library.path=$JAVA_LIBRARY_PATH"
  export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$JAVA_LIBRARY_PATH
fi  
HADOOP_OPTS="$HADOOP_OPTS -Dhadoop.policy.file=$HADOOP_POLICYFILE"
# Disable ipv6 as it can cause issues
HADOOP_OPTS="$HADOOP_OPTS -Djava.net.preferIPv4Stack=true"

一大波賦值語句就很少說了。

---

# put hdfs in classpath if present
if [ "$HADOOP_HDFS_HOME" = "" ]; then
  if [ -d "${HADOOP_PREFIX}/$HDFS_DIR" ]; then
    export HADOOP_HDFS_HOME=$HADOOP_PREFIX
  fi
fi

打印

HADOOP_HDFS_HOME=/root/hadoop-2.7.0-bin

---

if [ -d "$HADOOP_HDFS_HOME/$HDFS_DIR/webapps" ]; then
  CLASSPATH=${CLASSPATH}:$HADOOP_HDFS_HOME/$HDFS_DIR
fi
if [ -d "$HADOOP_HDFS_HOME/$HDFS_LIB_JARS_DIR" ]; then
  CLASSPATH=${CLASSPATH}:$HADOOP_HDFS_HOME/$HDFS_LIB_JARS_DIR'/*'
fi
CLASSPATH=${CLASSPATH}:$HADOOP_HDFS_HOME/$HDFS_DIR'/*'

打印

CLASSPATH=/root/hadoop-2.7.0-bin/etc/hadoop:/root/hadoop-2.7.0-bin/share/hadoop/common/lib/*:/root/hadoop-2.7.0-bin/share/hadoop/common/*:/root/hadoop-2.7.0-bin/share/hadoop/hdfs:/root/hadoop-2.7.0-bin/share/hadoop/hdfs/lib/*:/root/hadoop-2.7.0-bin/share/hadoop/hdfs/*

---

# put yarn in classpath if present
if [ "$HADOOP_YARN_HOME" = "" ]; then
  if [ -d "${HADOOP_PREFIX}/$YARN_DIR" ]; then
    export HADOOP_YARN_HOME=$HADOOP_PREFIX
  fi
fi

打印

HADOOP_YARN_HOME=/root/hadoop-2.7.0-bin

---

if [ -d "$HADOOP_YARN_HOME/$YARN_DIR/webapps" ]; then
  CLASSPATH=${CLASSPATH}:$HADOOP_YARN_HOME/$YARN_DIR
fi
if [ -d "$HADOOP_YARN_HOME/$YARN_LIB_JARS_DIR" ]; then
  CLASSPATH=${CLASSPATH}:$HADOOP_YARN_HOME/$YARN_LIB_JARS_DIR'/*'
fi
CLASSPATH=${CLASSPATH}:$HADOOP_YARN_HOME/$YARN_DIR'/*'
# put mapred in classpath if present AND different from YARN
if [ "$HADOOP_MAPRED_HOME" = "" ]; then
  if [ -d "${HADOOP_PREFIX}/$MAPRED_DIR" ]; then
    export HADOOP_MAPRED_HOME=$HADOOP_PREFIX
  fi
fi
if [ "$HADOOP_MAPRED_HOME/$MAPRED_DIR" != "$HADOOP_YARN_HOME/$YARN_DIR" ] ; then
  if [ -d "$HADOOP_MAPRED_HOME/$MAPRED_DIR/webapps" ]; then
    CLASSPATH=${CLASSPATH}:$HADOOP_MAPRED_HOME/$MAPRED_DIR
  fi
  if [ -d "$HADOOP_MAPRED_HOME/$MAPRED_LIB_JARS_DIR" ]; then
    CLASSPATH=${CLASSPATH}:$HADOOP_MAPRED_HOME/$MAPRED_LIB_JARS_DIR'/*'
  fi
  CLASSPATH=${CLASSPATH}:$HADOOP_MAPRED_HOME/$MAPRED_DIR'/*'
fi

打印

 CLASSPATH=/root/hadoop-2.7.0-bin/etc/hadoop:/root/hadoop-2.7.0-bin/share/hadoop/common/lib/*:/root/hadoop-2.7.0-bin/share/hadoop/common/*:/root/hadoop-2.7.0-bin/share/hadoop/hdfs:/root/hadoop-2.7.0-bin/share/hadoop/hdfs/lib/*:/root/hadoop-2.7.0-bin/share/hadoop/hdfs/*:/root/hadoop-2.7.0-bin/share/hadoop/yarn/lib/*:/root/hadoop-2.7.0-bin/share/hadoop/yarn/*:/root/hadoop-2.7.0-bin/share/hadoop/mapreduce/lib/*:/root/hadoop-2.7.0-bin/share/hadoop/mapreduce/*

---

# Add the user-specified CLASSPATH via HADOOP_CLASSPATH
# Add it first or last depending on if user has
# set env-var HADOOP_USER_CLASSPATH_FIRST
# if the user set HADOOP_USE_CLIENT_CLASSLOADER, HADOOP_CLASSPATH is not added
# to the classpath
if [[ ( "$HADOOP_CLASSPATH" != "" ) && ( "$HADOOP_USE_CLIENT_CLASSLOADER" = "" ) ]]; then
  # Prefix it if its to be preceded
  if [ "$HADOOP_USER_CLASSPATH_FIRST" != "" ]; then
    CLASSPATH=${HADOOP_CLASSPATH}:${CLASSPATH}
  else
    CLASSPATH=${CLASSPATH}:${HADOOP_CLASSPATH}
  fi
fi

打印

CLASSPATH=/root/hadoop-2.7.0-bin/etc/hadoop:/root/hadoop-2.7.0-bin/share/hadoop/common/lib/*:/root/hadoop-2.7.0-bin/share/hadoop/common/*:/root/hadoop-2.7.0-bin/share/hadoop/hdfs:/root/hadoop-2.7.0-bin/share/hadoop/hdfs/lib/*:/root/hadoop-2.7.0-bin/share/hadoop/hdfs/*:/root/hadoop-2.7.0-bin/share/hadoop/yarn/lib/*:/root/hadoop-2.7.0-bin/share/hadoop/yarn/*:/root/hadoop-2.7.0-bin/share/hadoop/mapreduce/lib/*:/root/hadoop-2.7.0-bin/share/hadoop/mapreduce/*:/contrib/capacity-scheduler/*.jar

---而後再回到最開始的hadoop腳本

|

|

|

---繼續執行

function print_usage(){
  echo "Usage: hadoop [--config confdir] [COMMAND | CLASSNAME]"
  echo "  CLASSNAME            run the class named CLASSNAME"
  echo " or"
  echo "  where COMMAND is one of:"
  echo "  fs                   run a generic filesystem user client"
  echo "  version              print the version"
  echo "  jar <jar>            run a jar file"
  echo "                       note: please use \"yarn jar\" to launch"
  echo "                             YARN applications, not this command."
  echo "  checknative [-a|-h]  check native hadoop and compression libraries availability"
  echo "  distcp <srcurl> <desturl> copy file or directories recursively"
  echo "  archive -archiveName NAME -p <parent path> <src>* <dest> create a hadoop archive"
  echo "  classpath            prints the class path needed to get the"
  echo "  credential           interact with credential providers"
  echo "                       Hadoop jar and the required libraries"
  echo "  daemonlog            get/set the log level for each daemon"
  echo "  trace                view and modify Hadoop tracing settings"
  echo ""
  echo "Most commands print help when invoked w/o parameters."
}

 這個就很少說了

很少細說

---

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

這個就不用我多說了。

---接下來開始執行命令

走的分支爲

 #hdfs commands
  namenode|secondarynamenode|datanode|dfs|dfsadmin|fsck|balancer|fetchdt|oiv|dfsgroups|portmap|nfs3)

    exit
    echo "DEPRECATED: Use of this script to execute hdfs command is deprecated." 1>&2
    echo "Instead use the hdfs command for it." 1>&2
    echo "" 1>&2
    #try to locate hdfs and if present, delegate to it.  
    shift
    if [ -f "${HADOOP_HDFS_HOME}"/bin/hdfs ]; then
      exec "${HADOOP_HDFS_HOME}"/bin/hdfs ${COMMAND/dfsgroups/groups}  "$@"
    elif [ -f "${HADOOP_PREFIX}"/bin/hdfs ]; then
      exec "${HADOOP_PREFIX}"/bin/hdfs ${COMMAND/dfsgroups/groups} "$@"
    else
      echo "HADOOP_HDFS_HOME not found!"
      exit 1
    fi
    ;;

最終執行的是

/root/hadoop-2.7.0-bin/bin/hdfs namenode -format

下一篇來說解這個命令。