当前位置：首页 > news >正文

hive迁移补数脚本细粒度表名-分区唯一键

news 2025/9/8 9:36:16

假设我通过对数脚本发现，这些表对不上。

假设检测出来是这样的（这些表存在于源端，但不存在目标端）

我们需要从源端迁移过去。

diff.txt

ads_xx1   dt=20250219
ads_xx2   dt=20250217
ads_xx2   dt=20250218
ads_xx2   dt=20250219
ads_xx3   dt=20250217
ads_xx4   dt=20250217
bak_xx1   dt=20250109
bak_xx1   dt=20250110
bak_xx1   dt=20250111
bak_xx1   dt=20250112
bak_xx1   dt=20250113
dim_e   dt=20250131
dim_j   dt=20250216
dim_j   dt=20250217
dim_m_df   dt=20250216
dim_m_df   dt=20250217
dim_m_df   dt=20250218
tmp_np
tmp_an   dt=20250217
tmp_an   dt=20250218
tmp_dhc_allcon_20250131
tmp_hj   dt=20250216
tmp_hj   dt=20250217
tmp_hj_2   dt=20250216
tmp_hj_2   dt=20250217

~

脚本

#!/bin/bash
#场景：数据在同一库下，并且hive是内部表（前缀的hdfs地址是相同的）




echo "" > rs.txt

#1.读取每一行，获取到表名和分区名
  
#定义map
declare -A table_map
while IFS=' ' read -r table_name fenqu_name
do
 #table_name为空的情况
     if [[ -z "${table_name// }" ]]; then
      echo "变量为空或只有空格，跳过操作"
      continue
     else
        #1.如果map中存在此表，并且分区不为空，则加入数组中
        if [[ -v table_map["$table_name"] ]]; then
          echo "$table_name exists in the map!"
          value_arr="${table_map[$table_name]}"
          IFS=' ' read -r -a new_arr <<< "${table_map["$table_name"]}"
          #判断分区为空吗
          if [ ${#new_arr[@]} -eq 0 ]; then
              echo "Array is empty!,不操作"
          else
              echo "Array is not empty!加入数组，放到map"
              new_arr+=("$fenqu_name")
              table_map["$table_name"]="$(IFS=' '; echo "${new_arr[*]}")"
          fi
        else
          #2.如果map中不存在此表，则创建数组，分区不可能存在又缺分区又缺表的存在（跑的时候筛出来,分批跑），则放进去。
          echo "$table_name does not exist in the map!"
          value_arr=()
          value_arr+=("$fenqu_name")
          table_map["$table_name"]="$(IFS=' '; echo "${value_arr[*]}")"
          

      fi


  fi

done < "$1"
echo "-----------"
#测试一下数组中的元素
#for table_name in "${!table_map[@]}"; do
#    echo "$table_name: ${table_map[$table_name]}"
#done



#2.遍历map3.遍历map
#删除每一个map里对应的分区
for table_name in "${!table_map[@]}"; do
    fenqu_arr_str=${table_map[$table_name]}
    echo "$table_name: $fenqu_arr_str"
    
    #如果没有分区，则删除表
    if [[ -z "$fenqu_arr_str" || "$fenqu_arr_str" =~ ^[[:space:]]*$ ]]; then
      echo "hdfs dfs -rm -r hdfs://xx.xx.xx.104:4007/apps/hive/warehouse/bigdata.db/$table_name"
      eval "hdfs dfs -rm -r hdfs://xx.xx.xx.104:4007/apps/hive/warehouse/bigdata.db/$table_name"
    # 在这里执行相应的操作
    else

      IFS=' ' read -r -a value_array <<< "$fenqu_arr_str"
      # 遍历数组
      for fenqu_name in "${value_array[@]}"; do
      # 输出每个 dt
        echo "hdfs dfs -rm -r hdfs://xx.xx.xx.104:4007/apps/hive/warehouse/bigdata.db/$table_name/$fenqu_name"
        eval "hdfs dfs -rm -r hdfs://xx.xx.xx.104:4007/apps/hive/warehouse/bigdata.db/$table_name/$fenqu_name"
      done


    fi  

done



#3.distcp任务
for table_name in "${!table_map[@]}"; do
    fenqu_arr_str=${table_map[$table_name]}
    echo "$table_name: $fenqu_arr_str"
    
    distcp_str="hadoop distcp -skipcrccheck -i -strategy dynamic -bandwidth 30 -m 20"
    source_path=hdfs://xx.xx.xx.7:8020/apps/hive/warehouse/bigdata.db/$table_name
    target_path=hdfs://xx.xx.xx.104:4007/apps/hive/warehouse/bigdata.db/$table_name

    #如果没有分区，则distcp表
    if [[ -z "$fenqu_arr_str" || "$fenqu_arr_str" =~ ^[[:space:]]*$ ]]; then
      distcp_str="$distcp_str $source_path $target_path"
      echo "$distcp_str"
      eval "$distcp_str"
    else

      IFS=' ' read -r -a value_array <<< "$fenqu_arr_str"
      # 遍历数组
      for fenqu_name in "${value_array[@]}"; do
        # 拼接
        distcp_str="$distcp_str $source_path/$fenqu_name"
      done
      #执行
      distcp_str="$distcp_str $target_path"
      echo "$distcp_str"
      eval "$distcp_str"
    fi
    echo "$table_name 完成" >> rs.txt
done

BEELINE_CMD="beeline -u 'jdbc:hive2://xx.xx.xx.104:7001/cfc;principal=hadoop/xx.xx.xx.104@TBDS-09T7KXLE'"
#4.修复分区
for table_name in "${!table_map[@]}"; do
    # 执行MSCK REPAIR TABLE命令
    echo "Repairing partitions for table: $table_name"
    $BEELINE_CMD -e "MSCK REPAIR TABLE $table_name;"

    if [ $? -eq 0 ]; then
        echo "Successfully repaired partitions for table: $table_name"
    else
        echo "Failed to repair partitions for table: $table_name"
    fi
done