淺談訂單重構之 MySQL 分庫分表實戰篇

Posted on by

一、目標

本文將完成如下目標:

  • 分表數量: 256    分庫數量: 4
  • 以用戶ID(user_id) 為數據庫分片Key
  • 最後測試訂單創建，更新，刪除, 單訂單號查詢，根據user_id查詢列表操作。

架構圖:

表結構如下:

CREATE TABLE `order_XXX` (
  `order_id` bigint(20) unsigned NOT NULL,
  `user_id` int(11) DEFAULT '0' COMMENT '訂單id',
  `status` int(11) DEFAULT '0' COMMENT '訂單狀態',
  `booking_date` datetime DEFAULT NULL,
  `create_time` datetime DEFAULT NULL,
  `update_time` datetime DEFAULT NULL,
  PRIMARY KEY (`order_id`),
  KEY `idx_user_id` (`user_id`),
  KEY `idx_bdate` (`booking_date`),
  KEY `idx_ctime` (`create_time`),
  KEY `idx_utime` (`update_time`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

註:  000<= XXX <= 255, 本文重點在於分庫分表實踐, 隻保留具有代表性字段，其它場景可以在此基礎上做改進。

全局唯一ID設計

要求:1.全局唯一 2:粗略有序 3:可反解出庫編號

  • 1bit + 39bit時間差 + 8bit機器號 + 8bit用戶編號(庫號) + 8bit自增序列
訂單號組成項 保留字段 毫秒級時間差 機器數 用戶編號(表編號) 自增序列
所占字節(單位bit) 1 39 8 8 8

單機最大QPS: 256000 使用壽命: 17年

二、環境準備

1、基本信息

項 版本 備註
SpringBoot 2.1.10.RELEASE
Mango 1.6.16 wiki地址:https://github.com/jfaster/mango
HikariCP 3.2.0
Mysql 5.7 測試使用docker一鍵搭建

2、數據庫環境準備

進入mysql: 

#主庫
 mysql -h 172.30.1.21 -uroot -pbytearch

#從庫
 mysql -h 172.30.1.31 -uroot -pbytearch

進入容器

#主
docker exec -it db_1_master /bin/bash

#從
docker exec -it db_1_slave /bin/bash

查看運行狀態

#主
docker exec db_1_master sh -c 'mysql -u root -pbytearch -e "SHOW MASTER STATUS \G"'
#從
docker exec db_1_slave sh -c 'mysql -u root -pbytearch -e "SHOW SLAVE STATUS \G"'

3、建庫 & 導入分表

（1）在mysql master實例分別建庫

172.30.1.21(   o rder_db_ 1) ,  172.30.1.22( order_db_2) ,

172.30.1.23( ord er_db_3) ,   172.30.1.24( order_db_4 )

（2）依次導入建表SQL 命令為

mysql -uroot -pbytearch -h172.30.1.21 order_db_1<fast-cloud-mysql-sharding/doc/sql/order_db_1.sql;
mysql -uroot -pbytearch -h172.30.1.22 order_db_2<fast-cloud-mysql-sharding/doc/sql/order_db_2.sql;
mysql -uroot -pbytearch -h172.30.1.23 order_db_3<fast-cloud-mysql-sharding/doc/sql/order_db_3.sql;
mysql -uroot -pbytearch -h172.30.1.24 order_db_4<fast-cloud-mysql-sharding/doc/sql/order_db_4.sql;

三、配置&實踐

1、pom文件  

     <!-- mango 分庫分表中間件 --> 
            <dependency>
                <groupId>org.jfaster</groupId>
                <artifactId>mango-spring-boot-starter</artifactId>
                <version>2.0.1</version>
            </dependency>
         
             <!-- 分佈式ID生成器 -->
            <dependency>
                <groupId>com.bytearch</groupId>
                <artifactId>fast-cloud-id-generator</artifactId>
                <version>${version}</version>
            </dependency>

            <!-- https://mvnrepository.com/artifact/mysql/mysql-connector-java -->
            <dependency>
                <groupId>mysql</groupId>
                <artifactId>mysql-connector-java</artifactId>
                <version>6.0.6</version>
            </dependency>

2、常量配置

package com.bytearch.fast.cloud.mysql.sharding.common;

/**
 * 分庫分表策略常用常量
 */
public class ShardingStrategyConstant {
    /**
     * database 邏輯名稱 ，真實庫名為 order_db_XXX
     */
    public static final String LOGIC_ORDER_DATABASE_NAME = "order_db";
    /**
     * 分表數 256，一旦確定不可更改
     */
    public static final int SHARDING_TABLE_NUM = 256;

    /**
     * 分庫數, 不建議更改， 可以更改，但是需要DBA遷移數據
     */
    public static final int SHARDING_DATABASE_NODE_NUM = 4;
}

3、yml 配置

4主4從數據庫配置， 這裡僅測試默認使用root用戶密碼，生產環境不建議使用root用戶。

mango:
  scan-package: com.bytearch.fast.cloud.mysql.sharding.dao
  datasources:
    - name: order_db_1
      master:
        driver-class-name: com.mysql.cj.jdbc.Driver
        jdbc-url: jdbc:mysql://172.30.1.21:3306/order_db_1?useUnicode=true&characterEncoding=utf8&autoReconnect=true&rewriteBatchedState&connectTimeout=1000&socketTimeout=5000&useSSL=false
        user-name: root
        password: bytearch
        maximum-pool-size: 10
        connection-timeout: 3000
      slaves:
        - driver-class-name: com.mysql.cj.jdbc.Driver
          jdbc-url: jdbc:mysql://172.30.1.31:3306/order_db_1?useUnicode=true&characterEncoding=utf8&autoReconnect=true&rewriteBatchedState&connectTimeout=1000&socketTimeout=5000&useSSL=false
          user-name: root
          password: bytearch
          maximum-pool-size: 10
          connection-timeout: 3000
    - name: order_db_2
      master:
        driver-class-name: com.mysql.cj.jdbc.Driver
        jdbc-url: jdbc:mysql://172.30.1.22:3306/order_db_2?useUnicode=true&characterEncoding=utf8&autoReconnect=true&rewriteBatchedState&connectTimeout=1000&socketTimeout=5000&useSSL=false
        user-name: root
        password: bytearch
        maximum-pool-size: 10
        connection-timeout: 3000
      slaves:
        - driver-class-name: com.mysql.cj.jdbc.Driver
          jdbc-url: jdbc:mysql://172.30.1.32:3306/order_db_2?useUnicode=true&characterEncoding=utf8&autoReconnect=true&rewriteBatchedState&connectTimeout=1000&socketTimeout=5000&useSSL=false
          user-name: root
          password: bytearch
          maximum-pool-size: 10
          connection-timeout: 3000
    - name: order_db_3
      master:
        driver-class-name: com.mysql.cj.jdbc.Driver
        jdbc-url: jdbc:mysql://172.30.1.23:3306/order_db_3?useUnicode=true&characterEncoding=utf8&autoReconnect=true&rewriteBatchedState&connectTimeout=1000&socketTimeout=5000&useSSL=false
        user-name: root
        password: bytearch
        maximum-pool-size: 10
        connection-timeout: 3000
      slaves:
        - driver-class-name: com.mysql.cj.jdbc.Driver
          jdbc-url: jdbc:mysql://172.30.1.33:3306/order_db_3?useUnicode=true&characterEncoding=utf8&autoReconnect=true&rewriteBatchedState&connectTimeout=1000&socketTimeout=5000&useSSL=false
          user-name: root
          password: bytearch
          maximum-pool-size: 10
          connection-timeout: 3000
    - name: order_db_4
      master:
        driver-class-name: com.mysql.cj.jdbc.Driver
        jdbc-url: jdbc:mysql://172.30.1.24:3306/order_db_4?useUnicode=true&characterEncoding=utf8&autoReconnect=true&rewriteBatchedState&connectTimeout=1000&socketTimeout=5000&useSSL=false
        user-name: root
        password: bytearch
        maximum-pool-size: 10
        connection-timeout: 3000
      slaves:
        - driver-class-name: com.mysql.cj.jdbc.Driver
          jdbc-url: jdbc:mysql://172.30.1.34:3306/order_db_4?useUnicode=true&characterEncoding=utf8&autoReconnect=true&rewriteBatchedState&connectTimeout=1000&socketTimeout=5000&useSSL=false
          user-name: root
          password: bytearch
          maximum-pool-size: 10
          connection-timeout: 300

4、分庫分表策略

1). 根據order_id為shardKey分庫分表策略

package com.bytearch.fast.cloud.mysql.sharding.strategy;

import com.bytearch.fast.cloud.mysql.sharding.common.ShardingStrategyConstant;
import com.bytearch.id.generator.IdEntity;
import com.bytearch.id.generator.SeqIdUtil;
import org.jfaster.mango.sharding.ShardingStrategy;

/**
 * 訂單號分庫分表策略
 */
public class OrderIdShardingStrategy implements ShardingStrategy<Long, Long> {
    @Override
    public String getDataSourceFactoryName(Long orderId) {
        if (orderId == null || orderId < 0L) {
            throw new IllegalArgumentException("order_id is invalid!");
        }
        IdEntity idEntity = SeqIdUtil.decodeId(orderId);
        if (idEntity.getExtraId() >= ShardingStrategyConstant.SHARDING_TABLE_NUM) {
            throw new IllegalArgumentException("sharding table Num is invalid, tableNum:" + idEntity.getExtraId());
        }
        //1. 計算步長
        int step = ShardingStrategyConstant.SHARDING_TABLE_NUM / ShardingStrategyConstant.SHARDING_DATABASE_NODE_NUM;
        //2. 計算出庫編號
        long dbNo = Math.floorDiv(idEntity.getExtraId(), step) + 1;
        //3. 返回數據源名
        return String.format("%s_%s", ShardingStrategyConstant.LOGIC_ORDER_DATABASE_NAME, dbNo);
    }

    @Override
    public String getTargetTable(String logicTableName, Long orderId) {
        if (orderId == null || orderId < 0L) {
            throw new IllegalArgumentException("order_id is invalid!");
        }
        IdEntity idEntity = SeqIdUtil.decodeId(orderId);
        if (idEntity.getExtraId() >= ShardingStrategyConstant.SHARDING_TABLE_NUM) {
            throw new IllegalArgumentException("sharding table Num is invalid, tableNum:" + idEntity.getExtraId());
        }
        // 基於約定，真實表名為 logicTableName_XXX， XXX不足三位補0
        return String.format("%s_%03d", logicTableName, idEntity.getExtraId());
    }
}

2). 根據user_id 為shardKey分庫分表策略

package com.bytearch.fast.cloud.mysql.sharding.strategy;

import com.bytearch.fast.cloud.mysql.sharding.common.ShardingStrategyConstant;
import org.jfaster.mango.sharding.ShardingStrategy;

/**
 * 指定分片KEY 分庫分表策略
 */
public class UserIdShardingStrategy implements ShardingStrategy<Integer, Integer> {

    @Override
    public String getDataSourceFactoryName(Integer userId) {
        //1. 計算步長 即單庫放得表數量
        int step = ShardingStrategyConstant.SHARDING_TABLE_NUM / ShardingStrategyConstant.SHARDING_DATABASE_NODE_NUM;
        //2. 計算出庫編號
        long dbNo = Math.floorDiv(userId % ShardingStrategyConstant.SHARDING_TABLE_NUM, step) + 1;
        //3. 返回數據源名
        return String.format("%s_%s", ShardingStrategyConstant.LOGIC_ORDER_DATABASE_NAME, dbNo);
    }

    @Override
    public String getTargetTable(String logicTableName, Integer userId) {
        // 基於約定，真實表名為 logicTableName_XXX， XXX不足三位補0
        return String.format("%s_%03d", logicTableName, userId % ShardingStrategyConstant.SHARDING_TABLE_NUM);
    }
}

5、dao層編寫

1). OrderPartitionByIdDao

package com.bytearch.fast.cloud.mysql.sharding.dao;

import com.bytearch.fast.cloud.mysql.sharding.common.ShardingStrategyConstant;
import com.bytearch.fast.cloud.mysql.sharding.pojo.entity.OrderEntity;
import com.bytearch.fast.cloud.mysql.sharding.strategy.OrderIdShardingStrategy;
import org.jfaster.mango.annotation.*;

@DB(name = ShardingStrategyConstant.LOGIC_ORDER_DATABASE_NAME, table = "order")
@Sharding(shardingStrategy = OrderIdShardingStrategy.class)
public interface OrderPartitionByIdDao {

    @SQL("INSERT INTO #table (order_id, user_id, status, booking_date, create_time, update_time) VALUES" +
            "(:orderId,:userId,:status,:bookingDate,:createTime,:updateTime)"
    )
    int insertOrder(@TableShardingBy("orderId") @DatabaseShardingBy("orderId") OrderEntity orderEntity);

    @SQL("UPDATE #table set update_time = now()" +
            "#if(:bookingDate != null),booking_date = :bookingDate #end " +
            "#if (:status != null), status = :status #end" +
            "WHERE order_id = :orderId"
    )
    int updateOrderByOrderId(@TableShardingBy("orderId") @DatabaseShardingBy("orderId") OrderEntity orderEntity);


    @SQL("SELECT * FROM #table WHERE order_id = :1")
    OrderEntity getOrderById(@TableShardingBy @DatabaseShardingBy Long orderId);

    @SQL("SELECT * FROM #table WHERE order_id = :1")
    @UseMaster
    OrderEntity getOrderByIdFromMaster(@TableShardingBy @DatabaseShardingBy Long orderId);

6、單元測試

@SpringBootTest(classes = {Application.class})
@RunWith(SpringJUnit4ClassRunner.class)
public class ShardingTest {
    @Autowired
    OrderPartitionByIdDao orderPartitionByIdDao;

    @Autowired
    OrderPartitionByUserIdDao orderPartitionByUserIdDao;

    @Test
    public void testCreateOrderRandom() {
        for (int i = 0; i < 20; i++) {
            int userId = ThreadLocalRandom.current().nextInt(1000,1000000);
            OrderEntity orderEntity = new OrderEntity();
            orderEntity.setOrderId(SeqIdUtil.nextId(userId % ShardingStrategyConstant.SHARDING_TABLE_NUM));
            orderEntity.setStatus(1);
            orderEntity.setUserId(userId);
            orderEntity.setCreateTime(new Date());
            orderEntity.setUpdateTime(new Date());
            orderEntity.setBookingDate(new Date());
            int ret = orderPartitionByIdDao.insertOrder(orderEntity);
            Assert.assertEquals(1, ret);
        }
    }

    @Test
    public void testOrderAll() {
        //insert
        int userId = ThreadLocalRandom.current().nextInt(1000,1000000);
        OrderEntity orderEntity = new OrderEntity();
        orderEntity.setOrderId(SeqIdUtil.nextId(userId % ShardingStrategyConstant.SHARDING_TABLE_NUM));
        orderEntity.setStatus(1);
        orderEntity.setUserId(userId);
        orderEntity.setCreateTime(new Date());
        orderEntity.setUpdateTime(new Date());
        orderEntity.setBookingDate(new Date());
        int i = orderPartitionByIdDao.insertOrder(orderEntity);
        Assert.assertEquals(1, i);

        //get from master
        OrderEntity orderInfo = orderPartitionByIdDao.getOrderByIdFromMaster(orderEntity.getOrderId());
        Assert.assertNotNull(orderInfo);
        Assert.assertEquals(orderInfo.getOrderId(), orderEntity.getOrderId());

        //get from slave
        OrderEntity slaveOrderInfo = orderPartitionByIdDao.getOrderById(orderEntity.getOrderId());
        Assert.assertNotNull(slaveOrderInfo);
        //update
        OrderEntity updateEntity = new OrderEntity();
        updateEntity.setOrderId(orderInfo.getOrderId());
        updateEntity.setStatus(2);
        updateEntity.setUpdateTime(new Date());
        int affectRows = orderPartitionByIdDao.updateOrderByOrderId(updateEntity);
        Assert.assertTrue( affectRows > 0);
    }

    @Test
    public void testGetListByUserId() {
        int userId = ThreadLocalRandom.current().nextInt(1000,1000000);
        for (int i = 0; i < 5; i++) {
            OrderEntity orderEntity = new OrderEntity();
            orderEntity.setOrderId(SeqIdUtil.nextId(userId % ShardingStrategyConstant.SHARDING_TABLE_NUM));
            orderEntity.setStatus(1);
            orderEntity.setUserId(userId);
            orderEntity.setCreateTime(new Date());
            orderEntity.setUpdateTime(new Date());
            orderEntity.setBookingDate(new Date());
            orderPartitionByIdDao.insertOrder(orderEntity);
        }
        try {
            //防止主從延遲引起的校驗錯誤
            Thread.sleep(1000);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        List<OrderEntity> orderListByUserId = orderPartitionByUserIdDao.getOrderListByUserId(userId);
        Assert.assertNotNull(orderListByUserId);
        Assert.assertTrue(orderListByUserId.size() == 5);
    }
}

大功告成:

四、總結

本篇主要介紹Java版使用Mango框架實現Mysql分庫分表實戰，分庫分表中間件也可以使用類似於ShardingJDBC，或者自研。

以上分庫分表數量僅供演示參考，實際工作中分表數量、分庫數量、是根據公司實際業務數據增長速度, 高峰期QPS,物理機器配置等等因素計算。

到此這篇關於淺談訂單重構之 MySQL 分庫分表實戰篇的文章就介紹到這瞭,更多相關MySQL 分庫分表內容請搜索WalkonNet以前的文章或繼續瀏覽下面的相關文章希望大傢以後多多支持WalkonNet！

推薦閱讀：