大数据常用shell脚本之ha-hadoop脚本

Posted 2021-12-08 广彐水厂

一、前置准备

CentOS7、jdk1.8、zookeeper-3.5.7、hadoop-2.7.7

想要完成本期视频中所有操作，需要以下准备：

二、简单了解

• 
  
• KaTeX parse error: Undefined control sequence: \\* at position 1: \\̲*̲表示传递给函数或脚本的所有参数…*”会将所有的参数作为一个整体
• 延时函数​​sleep​​​，默认以秒为单位，也可以自己指定延迟的单位。
  sleep 1s 延迟一秒
  sleep 1m 延迟一分钟
  sleep 1h 延迟一小时
  sleep 1d 延迟一天

三、脚本源码

#!/bin/bash



USAGE="使用方法：sh ha-hadoop.sh start/stop/status"
if [ $# -ne 1 ];then
  echo $USAGE
  exit 1
fi
SHELL_ZK=/home/xiaokang/bin/zk.sh
HADOOP_HOME=/opt/software/hadoop-2.7.7
NODES=("hadoop01" "hadoop02" "hadoop03")
case $1 in
"start")
  #启动zookeeper集群
  $SHELL_ZK start
  sleep 4s  
  #开始启动ha-hadoop集群
  for NODE in ${NODES[*]};do
    echo "--------$NODE启动ha-hadoop集群--------"
    if [ "hadoop01" = $NODE ];then
      ssh $NODE "$HADOOP_HOME/sbin/start-dfs.sh && $HADOOP_HOME/sbin/mr-jobhistory-daemon.sh start historyserver"   
    fi
    if [ "hadoop02" = $NODE ];then
      ssh $NODE "$HADOOP_HOME/sbin/yarn-daemon.sh start resourcemanager"  
    fi
    if [ "hadoop03" = $NODE ];then
      ssh $NODE "$HADOOP_HOME/sbin/start-yarn.sh" 
    fi
  done
  ;;
"stop")
  #开始停止ha-hadoop集群
  for NODE in ${NODES[*]};do
    echo "--------$NODE停止ha-hadoop集群--------"
    if [ "hadoop01" = $NODE ];then
      ssh $NODE "$HADOOP_HOME/sbin/stop-dfs.sh && $HADOOP_HOME/sbin/mr-jobhistory-daemon.sh stop historyserver"   
    fi
    if [ "hadoop02" = $NODE ];then
      ssh $NODE "$HADOOP_HOME/sbin/yarn-daemon.sh stop resourcemanager" 
    fi
    if [ "hadoop03" = $NODE ];then
      ssh $NODE "$HADOOP_HOME/sbin/stop-yarn.sh"  
    fi
  done
  #停止zookeeper集群
  $SHELL_ZK stop
  ;;
"status")
  for NODE in ${NODES[*]};do
    echo "--------$NODE查看ha-hadoop集群进程--------"
    ssh $NODE "jps" 
  done
  ;;
*)
  echo $USAGE 
  ;;
esac

四、测试使用

# 修改脚本权限
[xiaokang@hadoop01 bin]$ chmod 777 ha-hadoop.sh

执行方式：

# 方式一
[xiaokang@hadoop01 bin]$ ./ha-hadoop.sh start
# 方式二
[xiaokang@hadoop01 bin]$ sh ha-hadoop.sh start
# 方式三
# 1.将shell脚本所在目录配置为环境变量（/etc/profile），记得source生效
export SHELL_HOME=/home/xiaokang/bin
export PATH=${JAVA_HOME}/bin:${HADOOP_HOME}/bin:${HADOOP_HOME}/sbin:${ZOOKEEPER_HOME}/bin:${HIVE_HOME}/bin:${HBASE_HOME}/bin:${KYLIN_HOME}/bin:${KAFKA_HOME}/bin:${SCALA_HOME}/bin:${SCALA_HOME}/sbin:${SPARK_HOME}/bin:${SHELL_HOME}/:$PATH
# 2.在任意目录都可以使用自己编写的shell脚本
[xiaokang@hadoop01 ~]$ ha-hadoop.sh start

使用脚本进行启动和查看进程测试，成功如下图所示：

五、拓展脚本

查看集群进程脚本​​call-cluster.sh​​：

#!/bin/bash


USAGE="使用方法：sh call-cluster.sh jps or sh call-cluster.sh jps;java -version"
if [ $# -eq 0 ];then
        echo $USAGE
        exit 1
fi
NODES=("hadoop01" "hadoop02" "hadoop03")
for NODE in ${NODES[*]};do
        echo "--------$NODE--------"
        ssh $NODE "$*"
done
echo "----------------------------------------------------------------------------------------"
echo "--------c

不仅仅可以查看进程，还能同时完成好多事情，比如既查看集群进程也查看java版本：

[xiaokang@hadoop01 ~]$ call-cluster.sh jps;java -version