首页
登录 | 注册

HDFS数据迁移解决方案中DistCp工具的强大作用

#!/bin/bash
### nohup sh rc_CPData4.sh --copy 120 &

###避免同一个脚本重复启动
while [ $# -ne 2 ] || [ `ps x |grep -v grep|grep rc_CPData4.sh|wc -l` -gt 2 ] ; do
    echo "args are Incorrect"
    exit   
done

###配置需要同步的省份
#export HADOOP_USER_NAME="hdfs"
HADOOP_USER_NAME="hdfs"
declare -a provinces=("beijing" "shanghai" "guangdong")

###初始化并判断提交参数
command_type=$1
mapcount=$2
if ( [ $command_type != "--copy" ] || [ $command_type != "--update" ] ) && (( $mapcount <= 100 )) ; then
  echo "args are Incorrect,--copy or --updata and mapcount more than 100"
  exit
fi
DMRsetup="-Dmapreduce.job.queuename=ss_deploy -Dmapreduce.task.timeout=36000"
SOURCE_ROOT_DIR="hdfs://xxx:8020/serv/smartsteps/raw/events"
TARGET_ROOT_DIR="hdfs://xxx:8020/serv/smartsteps/raw/events"

###检测函数,判断源和目标是否存在
ret_Result()
{
  hadoop fs -test -e ${1}
  if [ $? -ne 0 ]; then echo  "failed";else echo "success";fi
}

###进入省循环体
for p in "${provinces[@]}"
do  
   ###读取需要同步的表,进入表循环
   for line in `cat ./list.txt | grep -v "#"`
   do  
       workspace=`echo $line |cut -d '|' -f1`
       table=`echo $line |cut -d '|' -f2`
   
 
       source_workspace=${SOURCE_ROOT_DIR}"/"${p}${workspace}
       source_table=${source_workspace}${table}

       target_workspace=${TARGET_ROOT_DIR}"/"${p}${workspace}
       target_table=${target_workspace}${table}
    
       source_table_flag=`ret_Result $source_table`  
           ###如果命令为:"--copy"并且源表存在就开始进入同步
       if  [ ${command_type} == "--copy" ] && [ ${source_table_flag} == "success" ];then
            target_workspace_flag=`ret_Result $target_workspace`
            target_table_flag=`ret_Result $target_table`

                        ###目标表存在就跳过
            if [ ${target_workspace_flag} == "success" ] && [ ${target_table_flag} == "success" ];then
                 echo "-----table already exist,skip copy --"${target_table}
            elif [ ${target_table_flag} == "failed" ];then
                 
                 if [ ${target_workspace_flag} == "failed" ];then
                     echo "-----creat workspace_dir" ${target_workspace}
                     hadoop fs -mkdir -p ${target_workspace}
                 else
                     echo "-----workspace_dir already exist" ${target_workspace}
                 fi
                 echo "-----start Synchronous data,table name:" ${table}
                 echo "-----from " ${source_table} "to" ${target_workspace}
                 hadoop distcp ${DMRsetup}  -m ${mapcount}  ${source_table}  ${target_workspace}
                                 ###判断同步结果是失败就删除临时文件,重新同步
                 if [ $? -ne 0 ] ; then
                      source_table_flag=`ret_Result $source_table`
                      target_table_flag=`ret_Result $target_table`
                      echo "-----Synchronous failed,del files :target_workspace/.distcp.tmp*,start updating it "
                      echo "-----start update table:" ${target_table}
                      hadoop fs -rm ${target_workspace}"/.distcp.tmp*"
                      if [ ${source_table_flag} == "success" ] && [ ${target_table_flag} == "success" ];then
                           hadoop distcp ${DMRsetup} -update -m ${mapcount}  ${source_table}  ${target_table}
                           if [ $? -ne 0 ]; then
                              echo "-----update failed,final------"
                              hadoop fs -rm ${target_table}"/.distcp.tmp*"
                           fi
                      elif [ ${source_table_flag} == "success" ] && [ ${target_table_flag} == "failed" ];then
                          echo "-----update table ,use Synchronous"
                          hadoop distcp ${DMRsetup} -m ${mapcount}  ${source_table}  ${target_workspace}
                           if [ $? -ne 0 ]; then
                              echo "-----update failed,final-----"
                              hadoop fs -rm ${target_workspace}"/.distcp.tmp*"
                           fi
                      fi
                 else
                      echo "=====Synchronous finished======"
                 fi

            fi
            ###如果命令为:"--update"并且源表存在就开始进入更新模块,
        elif  [ ${command_type} == "--update" ] && [ ${source_table_flag} == "success" ]; then        
           source_table_flag=`ret_Result $source_table`
           target_table_flag=`ret_Result $target_table`
           if [ ${source_table_flag} == "failed" ];then
               echo "-----source_table not exist-----"
           fi
           if [ ${target_table_flag} == "failed" ];then
               echo "-----target_table not exist-----"
           fi
                   
                   ###更新必须保持源表和目标表同时存在
           if [ ${source_table_flag} == "success" ] && [ ${target_table_flag} == "success" ];then
                 echo "start update data" ${source_table} "to" ${target_workspace}
                 hadoop distcp ${DMRsetup} -update -m ${mapcount}  ${source_table}  ${target_table}
                                 ###如果更新失败就删除临时文件
                 if [ $? -ne 0 ]; then
                     hadoop fs -rm ${target_table}"/.distcp.tmp*"
                     echo "-----update failed"
                 else
                     echo "-----update finished"  
                 fi        
           else
                 echo "-----"${target_table} "already not exist,can't update data----"
           fi
       else
          echo "-----source table not exist," $source_table
       fi   
       #echo $workspace
       #echo $source_workspace
       #echo $target_workspace       
   done
done


2020 jeepxie.net webmaster#jeepxie.net
10 q. 0.008 s.
京ICP备10005923号