在SpringBoot项目中使用多线程（配合线程池）加快从MySQL导入数据到ElasticSearch的速度

文章目录

• [1. 准备工作](#1. 准备工作)
• • [1.1 索引库](#1.1 索引库)
  • [1.2 建表](#1.2 建表)
  • [1.3 实体类](#1.3 实体类)
  • • [1.3.1 item.java](#1.3.1 item.java)
    • [1.3.2 itemDocument.java](#1.3.2 itemDocument.java)
  • [1.4 编写配置文件](#1.4 编写配置文件)
  • [1.5 编写 Mapper 类和 Service 类](#1.5 编写 Mapper 类和 Service 类)
• [2. 没有使用多线程的情况](#2. 没有使用多线程的情况)
• • [2.1 编码](#2.1 编码)
  • [2.2 测试结果](#2.2 测试结果)
• [3. 使用多线程（配合线程池）的情况](#3. 使用多线程（配合线程池）的情况)
• • [3.1 自定义类，实现 Runnable 接口](#3.1 自定义类，实现 Runnable 接口)
  • [3.2 编码（结合线程池）](#3.2 编码（结合线程池）)
  • [3.3 测试](#3.3 测试)
• [4. 对比及分析](#4. 对比及分析)

1. 准备工作

测试环境：

• JDK 17.0.7
• SpringBoot 3.0.2
• MySQL 8.0.34
• ElasticSearch 7.17.18

---

本次测试主要利用的是 Mybatis Plus、PageHelper、fastjson2

MybatisPlus 的 Maven 依赖

<dependency>
    <groupId>com.baomidou</groupId>
    <artifactId>mybatis-plus-boot-starter</artifactId>
    <version>3.5.7</version>
</dependency>

PageHelper 的 Maven 依赖

<dependency>
    <groupId>com.github.pagehelper</groupId>
    <artifactId>pagehelper-spring-boot-starter</artifactId>
    <version>2.1.0</version>
    <exclusions>
        <exclusion>
            <groupId>org.mybatis</groupId>
            <artifactId>mybatis</artifactId>
        </exclusion>
    </exclusions>
</dependency>

fastjson2的 Maven 依赖

<dependency>
    <groupId>com.alibaba.fastjson2</groupId>
    <artifactId>fastjson2</artifactId>
    <version>2.0.50</version>
</dependency>

1.1 索引库

创建一个名为 shopping_mall 的索引库

{
  "mappings": {
    "properties": {
      "id": {
        "type": "keyword"
      },
      "name": {
        "type": "text",
        "analyzer": "ik_smart"
      },
      "price": {
        "type": "integer"
      },
      "image": {
        "type": "keyword",
        "index": false
      },
      "category": {
        "type": "keyword"
      },
      "brand": {
        "type": "keyword"
      },
      "sold": {
        "type": "integer"
      },
      "commentCount": {
        "type": "integer",
        "index": false
      },
      "isAD": {
        "type": "boolean"
      },
      "updateTime": {
        "type": "date"
      }
    }
  }
}

1.2 建表

表名为 item，表结构如下（由于表中有 88476 条数据，无法在这里展示，需要具体的数据可以私聊我获取 SQL 文件）

/*
 Navicat Premium Data Transfer

 Source Server         : localhost
 Source Server Type    : MySQL
 Source Server Version : 80034 (8.0.34)
 Source Host           : localhost:3306
 Source Schema         : blog

 Target Server Type    : MySQL
 Target Server Version : 80034 (8.0.34)
 File Encoding         : 65001

 Date: 25/08/2024 01:59:24
*/

SET NAMES utf8mb4;
SET FOREIGN_KEY_CHECKS = 0;

-- ----------------------------
-- Table structure for item
-- ----------------------------
DROP TABLE IF EXISTS `item`;
CREATE TABLE `item`  (
  `id` bigint NOT NULL AUTO_INCREMENT COMMENT '商品id',
  `name` varchar(200) CHARACTER SET utf8mb3 COLLATE utf8mb3_general_ci NOT NULL COMMENT 'SKU名称',
  `price` int NOT NULL DEFAULT 0 COMMENT '价格（分）',
  `stock` int UNSIGNED NOT NULL COMMENT '库存数量',
  `image` varchar(200) CHARACTER SET utf8mb3 COLLATE utf8mb3_general_ci NULL DEFAULT NULL COMMENT '商品图片',
  `category` varchar(200) CHARACTER SET utf8mb3 COLLATE utf8mb3_general_ci NULL DEFAULT NULL COMMENT '类目名称',
  `brand` varchar(100) CHARACTER SET utf8mb3 COLLATE utf8mb3_general_ci NULL DEFAULT NULL COMMENT '品牌名称',
  `spec` varchar(200) CHARACTER SET utf8mb3 COLLATE utf8mb3_general_ci NULL DEFAULT NULL COMMENT '规格',
  `sold` int NULL DEFAULT 0 COMMENT '销量',
  `comment_count` int NULL DEFAULT 0 COMMENT '评论数',
  `isAD` tinyint(1) NULL DEFAULT 0 COMMENT '是否是推广广告，true/false',
  `status` int NULL DEFAULT 2 COMMENT '商品状态 1-正常，2-下架，3-删除',
  `create_time` datetime NULL DEFAULT CURRENT_TIMESTAMP COMMENT '创建时间',
  `update_time` datetime NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT '更新时间',
  `creater` bigint NULL DEFAULT NULL COMMENT '创建人',
  `updater` bigint NULL DEFAULT NULL COMMENT '修改人',
  PRIMARY KEY (`id`) USING BTREE,
  INDEX `status`(`status` ASC) USING BTREE,
  INDEX `updated`(`update_time` ASC) USING BTREE,
  INDEX `category`(`category` ASC) USING BTREE
) ENGINE = InnoDB AUTO_INCREMENT = 100002672305 CHARACTER SET = utf8mb3 COLLATE = utf8mb3_general_ci COMMENT = '商品表' ROW_FORMAT = COMPACT;

SET FOREIGN_KEY_CHECKS = 1;

1.3 实体类

1.3.1 item.java

import com.baomidou.mybatisplus.annotation.IdType;
import com.baomidou.mybatisplus.annotation.TableField;
import com.baomidou.mybatisplus.annotation.TableId;
import com.baomidou.mybatisplus.annotation.TableName;

import java.io.Serial;
import java.io.Serializable;
import java.time.LocalDateTime;

@TableName("item")
public class Item implements Serializable {

    @Serial
    private static final long serialVersionUID = 1L;

    /**
     * 商品id
     */
    @TableId(value = "id", type = IdType.AUTO)
    private Long id;

    /**
     * SKU名称
     */
    private String name;

    /**
     * 价格（分）
     */
    private Integer price;

    /**
     * 库存数量
     */
    private Integer stock;

    /**
     * 商品图片
     */
    private String image;

    /**
     * 类目名称
     */
    private String category;

    /**
     * 品牌名称
     */
    private String brand;

    /**
     * 规格
     */
    private String spec;

    /**
     * 销量
     */
    private Integer sold;

    /**
     * 评论数
     */
    private Integer commentCount;

    /**
     * 是否是推广广告，true/false
     */
    @TableField("isAD")
    private Boolean isAD;

    /**
     * 商品状态 1-正常，2-下架，3-删除
     */
    private Integer status;

    /**
     * 创建时间
     */
    private LocalDateTime createTime;

    /**
     * 更新时间
     */
    private LocalDateTime updateTime;

    /**
     * 创建人
     */
    private Long creater;

    /**
     * 修改人
     */
    private Long updater;

    public Long getId() {
        return id;
    }

    public void setId(Long id) {
        this.id = id;
    }

    public String getName() {
        return name;
    }

    public void setName(String name) {
        this.name = name;
    }

    public Integer getPrice() {
        return price;
    }

    public void setPrice(Integer price) {
        this.price = price;
    }

    public Integer getStock() {
        return stock;
    }

    public void setStock(Integer stock) {
        this.stock = stock;
    }

    public String getImage() {
        return image;
    }

    public void setImage(String image) {
        this.image = image;
    }

    public String getCategory() {
        return category;
    }

    public void setCategory(String category) {
        this.category = category;
    }

    public String getBrand() {
        return brand;
    }

    public void setBrand(String brand) {
        this.brand = brand;
    }

    public String getSpec() {
        return spec;
    }

    public void setSpec(String spec) {
        this.spec = spec;
    }

    public Integer getSold() {
        return sold;
    }

    public void setSold(Integer sold) {
        this.sold = sold;
    }

    public Integer getCommentCount() {
        return commentCount;
    }

    public void setCommentCount(Integer commentCount) {
        this.commentCount = commentCount;
    }

    public Boolean getIsAD() {
        return isAD;
    }

    public void setIsAD(Boolean AD) {
        isAD = AD;
    }

    public Integer getStatus() {
        return status;
    }

    public void setStatus(Integer status) {
        this.status = status;
    }

    public LocalDateTime getCreateTime() {
        return createTime;
    }

    public void setCreateTime(LocalDateTime createTime) {
        this.createTime = createTime;
    }

    public LocalDateTime getUpdateTime() {
        return updateTime;
    }

    public void setUpdateTime(LocalDateTime updateTime) {
        this.updateTime = updateTime;
    }

    public Long getCreater() {
        return creater;
    }

    public void setCreater(Long creater) {
        this.creater = creater;
    }

    public Long getUpdater() {
        return updater;
    }

    public void setUpdater(Long updater) {
        this.updater = updater;
    }

    @Override
    public String toString() {
        return "Item{" +
                "id=" + id +
                ", name='" + name + '\'' +
                ", price=" + price +
                ", stock=" + stock +
                ", image='" + image + '\'' +
                ", category='" + category + '\'' +
                ", brand='" + brand + '\'' +
                ", spec='" + spec + '\'' +
                ", sold=" + sold +
                ", commentCount=" + commentCount +
                ", isAD=" + isAD +
                ", status=" + status +
                ", createTime=" + createTime +
                ", updateTime=" + updateTime +
                ", creater=" + creater +
                ", updater=" + updater +
                '}';
    }

}

1.3.2 itemDocument.java

import java.time.LocalDateTime;

/**
 * 索引库实体类
 */
public class ItemDocument {
    
    private Long id;

    private String name;

    private Integer price;

    private Integer stock;

    private String image;

    private String category;

    private String brand;

    private Integer sold;

    private Integer commentCount;

    private Boolean isAD;

    private LocalDateTime updateTime;

    public Long getId() {
        return id;
    }

    public void setId(Long id) {
        this.id = id;
    }

    public String getName() {
        return name;
    }

    public void setName(String name) {
        this.name = name;
    }

    public Integer getPrice() {
        return price;
    }

    public void setPrice(Integer price) {
        this.price = price;
    }

    public Integer getStock() {
        return stock;
    }

    public void setStock(Integer stock) {
        this.stock = stock;
    }

    public String getImage() {
        return image;
    }

    public void setImage(String image) {
        this.image = image;
    }

    public String getCategory() {
        return category;
    }

    public void setCategory(String category) {
        this.category = category;
    }

    public String getBrand() {
        return brand;
    }

    public void setBrand(String brand) {
        this.brand = brand;
    }

    public Integer getSold() {
        return sold;
    }

    public void setSold(Integer sold) {
        this.sold = sold;
    }

    public Integer getCommentCount() {
        return commentCount;
    }

    public void setCommentCount(Integer commentCount) {
        this.commentCount = commentCount;
    }

    public Boolean getIsAD() {
        return isAD;
    }

    public void setIsAD(Boolean AD) {
        isAD = AD;
    }

    public LocalDateTime getUpdateTime() {
        return updateTime;
    }

    public void setUpdateTime(LocalDateTime updateTime) {
        this.updateTime = updateTime;
    }

    @Override
    public String toString() {
        return "ItemDocument{" +
                "id='" + id + '\'' +
                ", name='" + name + '\'' +
                ", price=" + price +
                ", stock=" + stock +
                ", image='" + image + '\'' +
                ", category='" + category + '\'' +
                ", brand='" + brand + '\'' +
                ", sold=" + sold +
                ", commentCount=" + commentCount +
                ", isAD=" + isAD +
                ", updateTime=" + updateTime +
                '}';
    }

}

1.4 编写配置文件

编写配置文件前，先导入 MySQL 连接驱动的 Maven 依赖

<dependency>
    <groupId>mysql</groupId>
    <artifactId>mysql-connector-java</artifactId>
</dependency>

application.yaml

spring:
  datasource:
    url: jdbc:mysql://localhost:3306/blog?useUnicode=true&characterEncoding=UTF-8&autoReconnect=true&serverTimezone=Asia/Shanghai&allowPublicKeyRetrieval=true
    driver-class-name: com.mysql.cj.jdbc.Driver
    username: root
    password: 123456

编写完配置文件后，在项目的启动类上添加 @MapperScan 注解，指定 Mapper 所在的包

@MapperScan("cn.edu.scau.mapper")

1.5 编写 Mapper 类和 Service 类

ItemMapper.java

import cn.edu.scau.pojo.Item;
import com.baomidou.mybatisplus.core.mapper.BaseMapper;

public interface ItemMapper extends BaseMapper<Item> {

}

ItemService.java

import cn.edu.scau.pojo.Item;
import com.baomidou.mybatisplus.extension.service.IService;

public interface ItemService extends IService<Item> {

}

ItemServiceImpl.java

import cn.edu.scau.mapper.ItemMapper;
import cn.edu.scau.pojo.Item;
import cn.edu.scau.service.ItemService;
import com.baomidou.mybatisplus.extension.service.impl.ServiceImpl;
import org.springframework.stereotype.Service;


@Service
public class ItemServiceImpl extends ServiceImpl<ItemMapper, Item> implements ItemService {

}

完成上述工作后，编写一个测试类，检查 ItemServiceImpl 类能否正常工作

import cn.edu.scau.service.ItemService;
import org.junit.jupiter.api.Test;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.boot.test.context.SpringBootTest;

@SpringBootTest
public class ItemServiceTests {

    @Autowired
    private ItemService itemService;

    @Test
    public void test() {
        System.out.println(itemService.getById(317578L));
    }

}

2. 没有使用多线程的情况

我们先来测试一下没有使用多线程的情况

2.1 编码

import cn.edu.scau.pojo.Item;
import cn.edu.scau.pojo.ItemDocument;
import cn.edu.scau.service.ItemService;
import com.alibaba.fastjson2.JSON;
import com.baomidou.mybatisplus.core.conditions.query.QueryWrapper;
import com.github.pagehelper.PageHelper;
import org.apache.http.HttpHost;
import org.elasticsearch.action.bulk.BulkRequest;
import org.elasticsearch.action.index.IndexRequest;
import org.elasticsearch.client.RequestOptions;
import org.elasticsearch.client.RestClient;
import org.elasticsearch.client.RestHighLevelClient;
import org.elasticsearch.xcontent.XContentType;
import org.junit.jupiter.api.AfterEach;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.api.Test;
import org.springframework.beans.BeanUtils;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.boot.test.context.SpringBootTest;

import java.util.List;


@SpringBootTest
public class BulkInsertDocumentTests {

    private RestHighLevelClient restHighLevelClient;

    @Autowired
    private ItemService itemService;

    @Test
    public void testBulkInsertDocument() throws Exception {
        int pageNumber = 1;
        int pageSize = 500;
        while (true) {
            // 1.准备文档数据
            QueryWrapper<Item> queryWrapper = new QueryWrapper<>();
            queryWrapper.lambda().eq(Item::getStatus, 1);
            PageHelper.startPage(pageNumber, pageSize);
            List<Item> itemList = itemService.list(queryWrapper);

            if (itemList == null || itemList.isEmpty()) {
                return;
            }

            // 2.准备 BulkRequest 对象
            BulkRequest bulkRequest = new BulkRequest();

            // 3.准备请求参数
            ItemDocument itemDocument;
            for (Item item : itemList) {
                itemDocument = new ItemDocument();
                BeanUtils.copyProperties(item, itemDocument);
                bulkRequest.add(new IndexRequest("shopping_mall")
                        .id(item.getId().toString())
                        .source(JSON.toJSONString(itemDocument), XContentType.JSON));
            }

            // 4.发送请求
            restHighLevelClient.bulk(bulkRequest, RequestOptions.DEFAULT);

            // 5.翻页
            pageNumber++;
        }
    }

    @BeforeEach
    public void setUp() {
        restHighLevelClient = new RestHighLevelClient(RestClient.builder(
                new HttpHost("127.0.0.1", 9200, "http")
        ));
    }

    @AfterEach
    public void tearDown() throws Exception {
        restHighLevelClient.close();
    }

}

2.2 测试结果

GET /shopping_mall/_count

共有 88475 条数据

1. 第一次导入耗时 36 秒 954 毫秒
2. 第二次导入耗时 38 秒 454 毫秒
3. 第三次导入耗时 38 秒 910 毫秒
4. 第四次导入耗时 40 秒 671毫秒
5. 第五次导入耗时 38 秒 958毫秒
6. 第六次导入耗时 38 秒 470毫秒

3. 使用多线程（配合线程池）的情况

3.1 自定义类，实现 Runnable 接口

import org.elasticsearch.action.bulk.BulkRequest;
import org.elasticsearch.client.RequestOptions;
import org.elasticsearch.client.RestHighLevelClient;

import java.io.IOException;
import java.util.concurrent.CountDownLatch;

public class InsertDocumentThread implements Runnable {

    private final RestHighLevelClient restHighLevelClient;

    private final BulkRequest bulkRequest;

    private final CountDownLatch countDownLatch;

    public InsertDocumentThread(RestHighLevelClient restHighLevelClient, BulkRequest bulkRequest, CountDownLatch countDownLatch) {
        this.restHighLevelClient = restHighLevelClient;
        this.bulkRequest = bulkRequest;
        this.countDownLatch = countDownLatch;
    }

    @Override
    public void run() {
        try {
            restHighLevelClient.bulk(bulkRequest, RequestOptions.DEFAULT);
            countDownLatch.countDown();
        } catch (IOException e) {
            throw new RuntimeException(e);
        }
    }

}

3.2 编码（结合线程池）

@Test
public void testBulkInsertDocumentWithMultipleThread() {
    int availableProcessors = Runtime.getRuntime().availableProcessors();
    LinkedBlockingQueue<Runnable> linkedBlockingQueue = new LinkedBlockingQueue<>(1000);
    ThreadPoolExecutor threadPoolExecutor = new ThreadPoolExecutor(
            2 * availableProcessors,
            4 * availableProcessors,
            100,
            TimeUnit.SECONDS,
            linkedBlockingQueue,
            Thread::new,
            new ThreadPoolExecutor.DiscardPolicy()
    );

    int pageNumber = 1;
    int pageSize = 500;
    long count = itemService.count() % pageSize == 0 ? itemService.count() / pageSize : itemService.count() / pageSize + 1;
    CountDownLatch countDownLatch = new CountDownLatch((int) count);

    long start = System.currentTimeMillis();
    while (true) {
        // 1.准备文档数据
        QueryWrapper<Item> queryWrapper = new QueryWrapper<>();
        queryWrapper.lambda().eq(Item::getStatus, 1);
        PageHelper.startPage(pageNumber, pageSize);
        List<Item> itemList = itemService.list(queryWrapper);

        if (itemList == null || itemList.isEmpty()) {
            break;
        }

        // 2.准备 BulkRequest 对象
        BulkRequest bulkRequest = new BulkRequest();

        // 3.准备请求参数
        ItemDocument itemDocument;
        for (Item item : itemList) {
            itemDocument = new ItemDocument();
            BeanUtils.copyProperties(item, itemDocument);
            bulkRequest.add(new IndexRequest("shopping_mall")
                    .id(item.getId().toString())
                    .source(JSON.toJSONString(itemDocument), XContentType.JSON));
        }

        // 4.发送请求
        InsertDocumentThread insertDocumentThread = new InsertDocumentThread(restHighLevelClient, bulkRequest, countDownLatch);
        threadPoolExecutor.submit(insertDocumentThread);

        // 5.翻页
        pageNumber++;
    }

    try {
        countDownLatch.await();
    } catch (InterruptedException e) {
        e.printStackTrace();
    }

    long end = System.currentTimeMillis();
    System.out.println("耗时：" + (end - start) / 1000 + "秒 " + (end - start) % 1000 + " 毫秒");
}

3.3 测试

DELETE /shopping_mall

我们先删除 shopping_mall 索引库，再次进行批量导入操作

---

GET /shopping_mall/_count

共导入 88475 条数据

1. 第一次导入耗时 30秒 657 毫秒
2. 第二次导入耗时 35 秒 200 毫秒
3. 第三次导入耗时 32 秒 265 毫秒
4. 第四次导入耗时 34 秒 11 毫秒
5. 第五次导入耗时 30 秒 778 毫秒
6. 第六次导入耗时 32 秒 861 毫秒

4. 对比及分析

通过对比可以发现，使用多线程从 MySQL 批量导入数据到 ElasticSearch，虽然速度提升了一点，但是不多，可能是因为以下原因：

1. 服务器的 CPU 核心数：我在做测试时，数据库用的是本地的，但 ElasticSearch 用的是云服务器，云服务器的 CPU 配置是 2 核，这也可能是导致使用多线程批量导入数据速度提升不明显的原因
2. I/O 密集型操作：Elasticsearch 的索引操作通常是 I/O 密集型的，这意味着瓶颈可能在于网络延迟和 Elasticsearch 服务器的响应时间，而不是 CPU 的处理能力，在这种情况下，增加线程数可能不会显著提高性能，因为 I/O 操作无法并行执行得更快
3. 网络带宽限制：网络带宽可能是瓶颈（我使用的云服务器的带宽是 6M），特别是在批量插入大量数据时