hudi

1.Hudi的介绍

2.Hudi环境的搭建

maven安装

在Centos系统上安装maven

wget  https://dlcdn.apache.org/maven/maven-3/3.8.6/binaries/apache-maven-3.8.6-bin.tar.gz

配置maven的环境变量

# 解压缩
tar -zxvf apache-maven-3.8.6-bin.tar.gz
# 配置maven环境
vim /etc/profile

export MAVEN_HOME=/home/wl/apache-maven-3.8.6-bin
export PATH=$MAVEN_HOME/bin:$PATH


source /etc/profile

Hudi安装

下载hudi的源码包

wget   https://dlcdn.apache.org/hudi/0.9.0/hudi-0.9.0.src.tgz

解压缩hudi源码包并编译hudi

mvn clean install -DskipTests -DskipITs -Dscala-2.12 -Dspark3

3.Hudi的使用

spark写入hudi

准备数据，resident.txt

{"username":"zhangsan","password":"123456","address":"翻斗家园1号街","ip":"192.182.11"}
{"username":"lisi","password":"www","address":"翻斗家园2号街","ip":"192.182.11"}
{"username":"wangwu","password":"23456","address":"翻斗家园3号街","ip":"192.182.11"}
{"username":"lili","password":"1333","address":"翻斗家园4号街","ip":"192.182.11"}
{"username":"xaiozhang","password":"5778","address":"翻斗家园1号街","ip":"192.182.11"}

启动spark-shell,指定环境

bin/spark-shell \
> --jars /home/wl/hudi-jar/hudi-spark3-bundle_2.12-0.9.0.jar,\
> /home/wl/hudi-jar/spark-avro_2.12-3.0.1.jar,/home/wl/hudi-jar/spark_unused-1.0.0.jar \
> --conf "spark.serializer=org.apache.spark.serializer.KryoSerializer"

resident.txt，写入hudi表

# 读取json文件转换成dataframe
val df=spark.read.json("/home/wl/data/resident.txt")
# 导入运行环境

import org.apache.hudi.QuickstartUtils._
import scala.collection.JavaConversions._
import org.apache.spark.sql.SaveMode._
import org.apache.hudi.DataSourceReadOptions._
import org.apache.hudi.DataSourceWriteOptions._
import org.apache.hudi.config.HoodieWriteConfig._

val tableName = "resident"
val basePath = "hdfs://192.168.160.2:8020/spark"

# 打开粘贴模式防止换行出错
# 将数据写入hudi表 保存至hdfs
:paste


df.write.
	format("hudi").
  mode(Overwrite).
  options(getQuickstartWriteConfigs).
  option(PRECOMBINE_FIELD.key(),"username").
  option(RECORDKEY_FIELD.key(),"password").
  option(PARTITIONPATH_FIELD.key(), "ip").
  option(TBL_NAME.key(), tableName).
  save(basePath)

# control+d

参数解释:

• Overwrite:写入覆盖

• getQuickstartWriteConfigs:设置写入/更新数据至Hudi时，Shuffle时分区数目

• PRECOMBINE_FIELD.key():数据合并时，依据主键字段

• RECORDKEY_FIELD.key():每个记录的唯一id

• PARTITIONPATH_FIELD.key()：分区字段

• TABLE_NAME:表名

• basePath:保存的路径

查看hdfs

spark读取hudi数据

找到hudi分区目录，一级分区就使用 一个 /*

val df=spark.read.format("hudi").load(basePath+"/*")

打印Schema，会发现多了几个属性，这些字段属于Hudi管理数据时使用的相关字段

 |-- _hoodie_commit_time: string (nullable = true)
 |-- _hoodie_commit_seqno: string (nullable = true)
 |-- _hoodie_record_key: string (nullable = true)
 |-- _hoodie_partition_path: string (nullable = true)
 |-- _hoodie_file_name: string (nullable = true)
 |-- address: string (nullable = true)
 |-- ip: string (nullable = true)
 |-- password: string (nullable = true)
 |-- username: string (nullable = true)

查询表中数据

df.show

idea中使用hudi-插入读取

环境准备

创建maven工程依赖如下

<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0"
         xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
         xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
    <modelVersion>4.0.0</modelVersion>

    <groupId>org.example</groupId>
    <artifactId>hudi</artifactId>
    <packaging>pom</packaging>
    <version>1.0-SNAPSHOT</version>
    <modules>
        <module>hudi-core</module>
    </modules>
    <properties>
        <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
        <encoding>UTF-8</encoding>
        <java.version>1.8</java.version>
        <maven.compiler.source>1.8</maven.compiler.source>
        <maven.compiler.target>1.8</maven.compiler.target>
        <hudi.version>0.9.0</hudi.version>
        <spark.version>3.1.2</spark.version>
        <hadoop.version>3.3.1</hadoop.version>
        <scala.binary.version>2.12</scala.binary.version>
    </properties>

    <dependencies>

        <dependency>
            <groupId>org.apache.hudi</groupId>
            <artifactId>hudi-spark3-bundle_${scala.binary.version}</artifactId>
            <version>${hudi.version}</version>
        </dependency>
        <dependency>
            <groupId>org.apache.spark</groupId>
            <artifactId>spark-avro_${scala.binary.version}</artifactId>
            <version>${spark.version}</version>
        </dependency>
        <!-- Spark Core 依赖 -->
        <dependency>
            <groupId>org.apache.spark</groupId>
            <artifactId>spark-core_${scala.binary.version}</artifactId>
            <version>${spark.version}</version>
        </dependency>
        <!-- Spark SQL 依赖 -->
        <dependency>
            <groupId>org.apache.spark</groupId>
            <artifactId>spark-sql_${scala.binary.version}</artifactId>
            <version>${spark.version}</version>
        </dependency>

        <!-- Hadoop Client 依赖 -->
        <dependency>
            <groupId>org.apache.hadoop</groupId>
            <artifactId>hadoop-client</artifactId>
            <version>${hadoop.version}</version>
        </dependency>
        <!-- spark-hive依赖 -->
        <dependency>
            <groupId>org.apache.spark</groupId>
            <artifactId>spark-hive_2.12</artifactId>
            <version>${spark.version}</version>
        </dependency>


    </dependencies>


</project>

添加scala插件,并将hadoop的 core-site.xml和hdfs-site.xml配置文件放到resources目录下，并将spark的log4j.properties.template文件放到resources目录下改名log4j.properties

代码实现

读取本地的json文件，并写入hudi和读取

package com.wl.hudi

import org.apache.spark.sql.SparkSession
import org.apache.hudi.QuickstartUtils._
import org.apache.hudi.DataSourceWriteOptions._
import org.apache.hudi.config.HoodieWriteConfig._

/**
 * @author 没有梦想的java菜鸟
 * @Date 创建时间：2022/7/19 上午11:39
 * @qq 2315290571
 * @Description ${description}
 */
object simpleExample {
  def main(args: Array[String]): Unit = {
    // 创建sparkSql运行环境
    val spark = SparkSession.builder()
      .appName("hudiExample")
      .master("local[*]")
      .config("spark.serializer", "org.apache.spark.serializer.KryoSerializer")
      .getOrCreate()
    // 读取本地json文件转换成dataFrame 
    val df = spark.read.json("file:" + "/Users/wangliang/Documents/ideaProject/hudi/hudi-core/src/main/resources/resident.txt")
    // 设置表名和存储路径
    val tableName = "resident"
    val basePath = "hdfs://192.168.160.2:8020/spark"
    // 写入hudi
    df.write.format("hudi")
      .mode("overwrite")
      .options(getQuickstartWriteConfigs)
      .option(PRECOMBINE_FIELD.key(), "username")
      .option(RECORDKEY_FIELD.key(), "password")
      .option(PARTITIONPATH_FIELD.key(), "ip")
      .option(TBL_NAME.key(), tableName)
      .save(basePath)
    // 读取hudi表
    spark.read.format("hudi").load(basePath+"/*").show()

    spark.stop()

  }
}

运行结果

idea中使用hudi-更新增量删除

更新数据

准备数据resident1.txt

{"username":"zhangsan","password":"123456","address":"翻斗家园2号街","ip":"192.182.11"}
{"username":"lisi","password":"www","address":"翻斗家园3号街","ip":"192.182.11"}
{"username":"wangwu","password":"23456","address":"翻斗家园4号街","ip":"192.182.11"}
{"username":"lili","password":"1333","address":"翻斗家园5号街","ip":"192.182.11"}
{"username":"xaiozhang","password":"5778","address":"翻斗家园6号街","ip":"192.182.11"}
{"username":"xiaowang","password":"333333","address":"翻斗家园6号街","ip":"192.182.11"}
{"username":"hututu","password":"1111111","address":"翻斗家园10号街","ip":"192.182.11"}

更新数据 只需要改动 写入文件方式，hudi会帮我们 upsert

package com.wl.hudi

import org.apache.hudi.DataSourceWriteOptions.{PARTITIONPATH_FIELD, PRECOMBINE_FIELD, RECORDKEY_FIELD}
import org.apache.hudi.QuickstartUtils.getQuickstartWriteConfigs
import org.apache.hudi.config.HoodieWriteConfig.TBL_NAME
import org.apache.spark.sql.{SaveMode, SparkSession}

/**
 * @author 没有梦想的java菜鸟
 * @Date 创建时间：2022/7/19 下午4:36
 * @qq 2315290571
 * @Description ${description}
 */
object updateExample {
  def main(args: Array[String]): Unit = {
     // 创建sparkSql运行环境
    val spark = SparkSession.builder()
      .appName("hudiExample")
      .master("local[*]")
      .config("spark.serializer", "org.apache.spark.serializer.KryoSerializer")
      .getOrCreate()
    // 读取json文件转换成dataFrame
    val df = spark.read.json("file:" + "/Users/wangliang/Documents/ideaProject/hudi/hudi-core/src/main/resources/resident1.txt")
    // 设置表名和存储路径
    val tableName = "resident"
    val basePath = "hdfs://192.168.160.2:8020/spark"

    // 更新数据
    df.write.format("hudi")
      .mode(SaveMode.Append)
      .options(getQuickstartWriteConfigs)
      .option(PRECOMBINE_FIELD.key(), "username")
      .option(RECORDKEY_FIELD.key(), "password")
      .option(PARTITIONPATH_FIELD.key(), "ip")
      .option(TBL_NAME.key(), tableName)
      .save(basePath)

    // 查询更新后的数据
    // 读取hudi表
    spark.read.format("hudi").load(basePath+"/*").show()
  }

}

运行结果

增量查询

当hudi表类型是COW时候，支持 快照查询(Snapshot Queries)和增量查询(Incremental Queries) ，默认情况下是快照查询。我们可以通过option 进行设置，最重要的两个参数如下图

设置查询类型

设置开始时间

package com.wl.hudi

import org.apache.spark.sql.{Row, SparkSession}
import org.apache.hudi.DataSourceReadOptions._

/**
 * @author 没有梦想的java菜鸟
 * @Date 创建时间：2022/7/19 下午5:29
 * @qq 2315290571
 * @Description ${description}
 */
object IncrementExample {
  def main(args: Array[String]): Unit = {
    // 创建sparkSql运行环境
    val spark = SparkSession.builder()
      .appName("hudiExample")
      .master("local[*]")
      .config("spark.serializer", "org.apache.spark.serializer.KryoSerializer")
      .getOrCreate()
    // 设置表名和存储路径
    val basePath = "hdfs://192.168.160.2:8020/spark"

    import spark.implicits._
    // 加载hudi数据 注册临时表 并  获取  hoodie_commit_time 当做 beginTime
    spark.read
      .format("hudi")
      .load(basePath + "/*")
      .createOrReplaceTempView("resident")

    val rows: Array[String] = spark.sql(
      """
        |select
        |  distinct(_hoodie_commit_time) as commitTime
        |from
        |  resident
        |order by
        |   commitTime DESC
        |
        |""".stripMargin).map(k => k.getString(0)).take(10)
    // 因为我的时间都是同一天  所以需要转换一下
    val beginTime = rows(rows.length - 1).replaceAll("20220719","20220718")
    println(s"beginTime = ${beginTime}")
    // 增量查询
    val df = spark.read
      .format("hudi")
      // 设置查询数据模式为:incremental,增量读取
      .option(QUERY_TYPE.key(), QUERY_TYPE_INCREMENTAL_OPT_VAL)
      // 设置增量读取数据的开始时间  >beginTime
      .option(BEGIN_INSTANTTIME.key(), beginTime)
      .load(basePath)
    // 注册临时表并查询结果
    df.createOrReplaceTempView("resident_incremental")

    spark.sql("select * from resident_incremental").show
  }
}

运行结果

+-------------------+--------------------+------------------+----------------------+--------------------+--------------+----------+--------+---------+
|_hoodie_commit_time|_hoodie_commit_seqno|_hoodie_record_key|_hoodie_partition_path|   _hoodie_file_name|       address|        ip|password| username|
+-------------------+--------------------+------------------+----------------------+--------------------+--------------+----------+--------+---------+
|     20220719165757|  20220719165757_0_1|              1333|            192.182.11|a9aab30f-3263-4cc...| 翻斗家园5号街|192.182.11|    1333|     lili|
|     20220719165757|  20220719165757_0_2|             23456|            192.182.11|a9aab30f-3263-4cc...| 翻斗家园4号街|192.182.11|   23456|   wangwu|
|     20220719165757|  20220719165757_0_3|               www|            192.182.11|a9aab30f-3263-4cc...| 翻斗家园3号街|192.182.11|     www|     lisi|
|     20220719165757|  20220719165757_0_4|              5778|            192.182.11|a9aab30f-3263-4cc...| 翻斗家园6号街|192.182.11|    5778|xaiozhang|
|     20220719165757|  20220719165757_0_5|            123456|            192.182.11|a9aab30f-3263-4cc...| 翻斗家园2号街|192.182.11|  123456| zhangsan|
|     20220719165757|  20220719165757_0_6|           1111111|            192.182.11|a9aab30f-3263-4cc...|翻斗家园10号街|192.182.11| 1111111|   hututu|
|     20220719165757|  20220719165757_0_7|            333333|            192.182.11|a9aab30f-3263-4cc...| 翻斗家园6号街|192.182.11|  333333| xiaowang|
+-------------------+--------------------+------------------+----------------------+--------------------+--------------+----------+--------+---------+

删除数据

准备用于删除的json文件,resident2.txt

{"username":"zhangsan","password":"123456","address":"翻斗家园2号街","ip":"192.182.11"}
{"username":"lisi","password":"www","address":"翻斗家园3号街","ip":"192.182.11"}

代码

package com.wl.hudi

import org.apache.spark.sql.{SaveMode, SparkSession}

import org.apache.hudi.QuickstartUtils.getQuickstartWriteConfigs
import org.apache.hudi.DataSourceWriteOptions._
import org.apache.hudi.config.HoodieWriteConfig._

/**
 * @author 没有梦想的java菜鸟
 * @Date 创建时间：2022/7/19 下午6:55
 * @qq 2315290571
 * @Description ${description}
 */
object DeleteExample {
  def main(args: Array[String]): Unit = {
    // 创建sparkSql运行环境
    val spark = SparkSession.builder()
      .appName("hudiExample")
      .master("local[*]")
      .config("spark.serializer", "org.apache.spark.serializer.KryoSerializer")
      .getOrCreate()
    // 设置表名和存储路径
    val basePath = "hdfs://192.168.160.2:8020/spark"
    val tableName = "resident"
    // 读取删除的数据
    val df = spark.read.json("file:" + "/Users/wangliang/Documents/ideaProject/hudi/hudi-core/src/main/resources/resident2.txt")
    // 删除语句
    df.write
      .mode(SaveMode.Append)
      .format("hudi")
      .options(getQuickstartWriteConfigs())
       // 指定写入 方式 是删除
      .option(OPERATION.key(), "delete")
      .option(PRECOMBINE_FIELD.key(), "username")
      .option(RECORDKEY_FIELD.key(), "password")
      .option(PARTITIONPATH_FIELD.key(), "ip")
      .option(TBL_NAME.key(), tableName)
      .save(basePath)

    // 再查表看看是否减少了数据
     spark.read.format("hudi").load(basePath+"/*").show

    spark.stop()

  }

}

查询结果