二百四十五、海豚调度器——用DolphinScheduler调度执行复杂的HiveSQL(HQL包含多种海豚无法正确识别的符号)

天地风雷水火山泽2024-07-25 17:08

一、目的

在Hive中完成复杂JSON，既有对象还有数组而且数组中包含数组的解析后，原本以为没啥问题了，结果在DolphinScheduler中调度又出现了大问题，搞了一天、试了很多种方法、死了无数脑细胞，才解决了这个问题！

二、HiveSQL

复制代码 
insert  overwrite  table  hurys_dc_dwd.dwd_json_statistics partition(day)
select
        t1.device_no,
        source_device_type,
        sn,
        model,
        create_time,
        cycle,
        get_json_object(coil_list,'$.laneNo')  lane_no,
        get_json_object(coil_list,'$.laneType')           lane_type,
        section_no,
        get_json_object(coil_list,'$.coilNo')             coil_no,
        get_json_object(coil_list,'$.volumeSum')          volume_sum,
        get_json_object(coil_list,'$.volumePerson')       volume_person,
        get_json_object(coil_list,'$.volumeCarNon')       volume_car_non,
        get_json_object(coil_list,'$.volumeCarSmall')     volume_car_small,
        get_json_object(coil_list,'$.volumeCarMiddle')    volume_car_middle,
        get_json_object(coil_list,'$.volumeCarBig')       volume_car_big,
        get_json_object(coil_list,'$.speedAvg')           speed_avg,
        get_json_object(coil_list,'$.speed85')            speed_85,
        get_json_object(coil_list,'$.timeOccupancy')      time_occupancy,
        get_json_object(coil_list,'$.averageHeadway')     average_headway,
        get_json_object(coil_list,'$.averageGap')         average_gap,
        substr(create_time,1,10) day
from (select
       get_json_object(statistics_json,'$.deviceNo')          device_no,
       get_json_object(statistics_json,'$.sourceDeviceType')  source_device_type,
       get_json_object(statistics_json,'$.sn')                sn,
       get_json_object(statistics_json,'$.model')             model,
       get_json_object(statistics_json,'$.createTime')        create_time ,
       get_json_object(statistics_json,'$.data.cycle')        cycle,
       get_json_object(replace(replace(section_list,':{',':[{'),'}}','}]}'),'$.sectionNo') section_no,
       section_list
from hurys_dc_ods.ods_statistics
lateral view explode(split(replace(replace(replace(get_json_object(statistics_json,'$.data.sectionList'),
    '[',''),']',''),'},{"sectionNo"','}|{"sectionNo"'),"\\|")) tf as section_list
    where day='2024-07-18' --  date_sub(current_date(), 1)   -- '2024-07-18' --
    ) as t1
lateral view explode(split(replace(replace(replace(get_json_object(replace(replace(section_list,
    ':{',':[{'),'}}','}]}'),'$.coilList'),'[',''),']',''),'},','}|'),"\\|")) tf1 as coil_list
    where substr(create_time,1,10) =  '2024-07-18' --date_sub(current_date(), 1)   --'2024-07-17'
;

三、原先海豚任务的调度方式

在shell脚本里添加HiveSQL语句

#! /bin/bash

source /etc/profile

nowdate=`date --date='0 days ago' "+%Y%m%d"`

yesdate=`date -d yesterday +%Y-%m-%d`

hive -e "

use hurys_dc_dwd;

set hive.vectorized.execution.enabled=false;

set hive.exec.dynamic.partition=true;

set hive.exec.dynamic.partition.mode=nonstrict;

set hive.exec.max.dynamic.partitions.pernode=1000;

set hive.exec.max.dynamic.partitions=1500;

with t1 as(

select

get_json_object(statistics_json,'$.deviceNo') device_no,

get_json_object(statistics_json,'$.sourceDeviceType') source_device_type,

get_json_object(statistics_json,'$.sn') sn,

get_json_object(statistics_json,'$.model') model,

get_json_object(statistics_json,'$.createTime') create_time ,

get_json_object(statistics_json,'$.data.cycle') cycle,

get_json_object(replace(replace(section_list,':{',':{'),'}}','}}'),'$.sectionNo') section_no,

section_list

from hurys_dc_ods.ods_statistics

lateral view explode(split(replace(replace(replace(get_json_object(statistics_json,'$.data.sectionList'),'',''),'',''),'},{"sectionNo"','}|{"sectionNo"'),"\\\\|")) tf as section_list

where day='$yesdate'

)

insert overwrite table hurys_dc_dwd.dwd_json_statistics partition(day)

select

t1.device_no,

source_device_type,

sn,

model,

substr(create_time,1,19) create_time ,

cycle,

get_json_object(coil_list,'$.laneNo') lane_no,

get_json_object(coil_list,'$.laneType') lane_type,

section_no,

get_json_object(coil_list,'$.coilNo') coil_no,

get_json_object(coil_list,'$.volumeSum') volume_sum,

get_json_object(coil_list,'$.volumePerson') volume_person,

get_json_object(coil_list,'$.volumeCarNon') volume_car_non,

get_json_object(coil_list,'$.volumeCarSmall') volume_car_small,

get_json_object(coil_list,'$.volumeCarMiddle') volume_car_middle,

get_json_object(coil_list,'$.volumeCarBig') volume_car_big,

get_json_object(coil_list,'$.speedAvg') speed_avg,

get_json_object(coil_list,'$.speed85') speed_85,

get_json_object(coil_list,'$.timeOccupancy') time_occupancy,

get_json_object(coil_list,'$.averageHeadway') average_headway,

get_json_object(coil_list,'$.averageGap') average_gap,

substr(create_time,1,10) day

from t1

lateral view explode(split(replace(replace(replace(get_json_object(replace(replace(section_list,':{',':{'),'}}','}}'),'$.coilList'),'',''),'',''),'},','}|'),"\\\\|")) tf1 as coil_list

where substr(create_time,1,10) ='$yesdate'

"

四、原先方式报错日志

海豚无法正确识别HiveSQL里解析复杂JSON的多种符号

五、解决方式

把HiveSQL放在一个SQL文件里，然后在脚本里是执行Hive的sourceSQL文件

1 SQL文件

--使用hurys_dc_ods数据库

use hurys_dc_dwd;

--hive调优（必须先执行调优语句，否则部分复杂SQL运行会有问题）

set hive.vectorized.execution.enabled=false;

--开启动态分区功能（默认 true，开启）

set hive.exec.dynamic.partition=true;

--设置为非严格模式 nonstrict 模式表示允许所有的分区字段都可以使用动态分区

set hive.exec.dynamic.partition.mode=nonstrict;

--在每个执行 MR 的节点上，最大可以创建多少个动态分区

set hive.exec.max.dynamic.partitions.pernode=1000;

--在所有执行 MR 的节点上，最大一共可以创建多少个动态分区。默认 1000

set hive.exec.max.dynamic.partitions=1500;

insert overwrite table hurys_dc_dwd.dwd_json_statistics partition(day)

select

t1.device_no,

source_device_type,

sn,

model,

create_time,

cycle,

get_json_object(coil_list,'$.laneNo') lane_no,

get_json_object(coil_list,'$.laneType') lane_type,

section_no,

get_json_object(coil_list,'$.coilNo') coil_no,

get_json_object(coil_list,'$.volumeSum') volume_sum,

get_json_object(coil_list,'$.volumePerson') volume_person,

get_json_object(coil_list,'$.volumeCarNon') volume_car_non,

get_json_object(coil_list,'$.volumeCarSmall') volume_car_small,

get_json_object(coil_list,'$.volumeCarMiddle') volume_car_middle,

get_json_object(coil_list,'$.volumeCarBig') volume_car_big,

get_json_object(coil_list,'$.speedAvg') speed_avg,

get_json_object(coil_list,'$.speed85') speed_85,

get_json_object(coil_list,'$.timeOccupancy') time_occupancy,

get_json_object(coil_list,'$.averageHeadway') average_headway,

get_json_object(coil_list,'$.averageGap') average_gap,

substr(create_time,1,10) day

from (select

get_json_object(statistics_json,'$.deviceNo') device_no,

get_json_object(statistics_json,'$.sourceDeviceType') source_device_type,

get_json_object(statistics_json,'$.sn') sn,

get_json_object(statistics_json,'$.model') model,

get_json_object(statistics_json,'$.createTime') create_time ,

get_json_object(statistics_json,'$.data.cycle') cycle,

get_json_object(replace(replace(section_list,':{',':{'),'}}','}}'),'$.sectionNo') section_no,

section_list

from hurys_dc_ods.ods_statistics

lateral view explode(split(replace(replace(replace(get_json_object(statistics_json,'$.data.sectionList'),'',''),'',''),'},{"sectionNo"','}|{"sectionNo"'),"\\|")) tf as section_list
where day= date_sub(current_date(), 1)

) as t1

lateral view explode(split(replace(replace(replace(get_json_object(replace(replace(section_list,':{',':{'),'}}','}}'),'$.coilList'),'',''),'',''),'},','}|'),"\\|")) tf1 as coil_list
where substr(create_time,1,10) = date_sub(current_date(), 1)

;

2 海豚任务执行脚本

#! /bin/bash

source /etc/profile

nowdate=`date --date='0 days ago' "+%Y-%m-%d"`

yesdate=`date -d yesterday +%Y-%m-%d`

hive -e "

source dwd_json_statistics.sql

"

3 执行任务，验证结果

终于解决了，以后碰到类似调度器识别不了SQL里符号的问题，实在不行就用这个方法，把SQL放在SQL文件里，然后在脚本里执行这个SQL文件就行了，这样就能规避这类问题了

相关推荐
SelectDB10 小时前
秒级弹性、最高降本 70%：SelectDB Serverless 如何重塑云数仓资源效率
大数据·后端·云原生
WhoAmI10 小时前
MapReduce框架原理解析一：InputFormat
大数据·hadoop
WhoAmI10 小时前
MapReduce框架原理解析三：OutputFormat
大数据·hadoop
WhoAmI10 小时前
MapReduce框架原理解析二：Shuffle
大数据·hadoop
大大大大晴天1 天前
Hudi技术内幕：Key Generation原理与实践
大数据
得物技术5 天前
从埋点需求到规则资产：Hermes Agent 重构得物数仓工作流
大数据·llm·ai编程
久美子5 天前
AI驱动数仓建设的Harness工程实践——本体建模、知识分层与上下文工程
大数据
大树885 天前
金刚石散热越强，管路越先见顶
大数据·运维·服务器·人工智能·ai
大志哥1235 天前
ES和Logstash日志链路系统上线后遭遇切片爆炸（解决）
大数据·elasticsearch
果丁智能6 天前
物联网智能锁赋能集中式住宿：身份核验与远程权限管控的全链路技术实践
大数据·人工智能·物联网·智能家居
热门推荐
012026年6月AI大模型全景报告：GPT-5.6、Claude Opus 4.8、Gemini 3.5，中美AI三足鼎立谁主沉浮？022026年6月AI行业全景：从百模大战到Agent元年，这30天发生了什么？032026 年 AI 编程工具终极横评：Cursor vs Claude Code vs Copilot vs Windsurf04【AI】2026 年具身智能模型和世界模型总结05GitHub 镜像站点06Claude Code、Codex、Cursor三分天下：2026年AI编程Agent生态全景剖析07AI一周事件 · 2026-06-03 至 2026-06-0908AI科技热点日报 | 2026年6月1日09Codex 下载安装指南：Windows 和 macOS 官方版下载10上线仅72小时被强制下架：Claude Fable 5 的短命