因資料量過大，執行sqoop跑不動或者卡記憶體，於是通過寫指令碼分批匯入到HDFS，然後再載入到Hive表中。
shell指令碼如下：

#!/bin/bash
source /etc/profile

host=127.0.0.1

for((i=1; i<=100; i++))
do   
    start=$(((${i} - 1) * 100000 + 1))
    end=$((${i} * 100000))

    sql="select person_id,capture_time,write_time,capture_resource_id,major_capture_image_url,minor_capture_image_url,sex,age,orientation,glasses,knapsack, bag,messenger_bag,shoulder_bag,umbrella,hair,hat,mask,upper_color,upper_type,upper_texture,bottom_color,bottom_type,trolley_case,barrow,baby,feature_type,feature_code from big_data.pedestrian_sm where person_id>=${start}
 
 and person_id<=${end} and \$CONDITIONS";  

    sqoop import --connect jdbc:mysql://${host}:3306/big_data \
    --username root \
    --password 123456 \
    --query "${sql}" \
    --fields-terminated-by '\001' \
    --delete-target-dir \
    --target-dir hdfs://hsmaster:9000/tmp/big_data/pedestrian_sm/${start}
 
-${end}/ \
    --split-by person_id \
    -m 8

    echo Sqoop import from: ${start} to: ${end} success....................................


    hive -e "
        use big_data;
        load data inpath 'hdfs://master:9000/tmp/big_data/pedestrian_sm/${start}-${end}' into table big_data.pedestrian_sm;
    "

    echo
 
 Hive load from: ${start}-${end} success....................................

done