原创

【Spring Boot 实战】数据库千万级分库分表和读写分离实战

一. 前言

前几天时间写了如何使用Sharding-JDBC进行分库分表和读写分离的例子,相信能够感受到Sharding-JDBC的强大了,而且使用配置都非常干净。官方支持的功能还很多功能分布式主键、强制路由等。这里是最终版介绍下如何在分库分表的基础上集成读写分离的功能。

推荐先阅读:
SpringBoot 2.x ShardingSphere分库分表实战
SpringBoot 2.x ShardingSphere读写分离实战

二. 项目实战

主从数据库配置

在配置前,我们希望分库分表规则和之前保持一致:

基于user表,根据id进行分库,如果id mod 2为奇数则落在ds0库,偶数则落在ds1库
根据age进行分表,如果age mod 2为奇数则落在user_0表,偶数则落在user_1表

读写分离规则:

读都落在从库,写落在主库

因为使用我们使用Sharding-JDBC Spring Boot Starter,所以还是只需要在properties配置文件配置主从库的数据源即可

# 可以看到配置四个数据源 分别是 主数据库两个 从数据库两个
sharding.jdbc.datasource.names=master0,master1,master0slave0,master1slave0
# 主第一个数据库
sharding.jdbc.datasource.master0.type=com.zaxxer.hikari.HikariDataSource
sharding.jdbc.datasource.master0.hikari.driver-class-name=com.mysql.jdbc.Driver
sharding.jdbc.datasource.master0.jdbc-url=jdbc:mysql://192.168.0.4:3306/ds0?characterEncoding=utf-8&serverTimezone=Asia/Shanghai
sharding.jdbc.datasource.master0.username=test
sharding.jdbc.datasource.master0.password=12root
# 主第二个数据库
sharding.jdbc.datasource.master1.type=com.zaxxer.hikari.HikariDataSource
sharding.jdbc.datasource.master1.hikari.driver-class-name=com.mysql.jdbc.Driver
sharding.jdbc.datasource.master1.jdbc-url=jdbc:mysql://192.168.0.4:3306/ds1?characterEncoding=utf-8&serverTimezone=Asia/Shanghai
sharding.jdbc.datasource.master1.username=test
sharding.jdbc.datasource.master1.password=12root
# 从第一个数据库
sharding.jdbc.datasource.master0slave0.type=com.zaxxer.hikari.HikariDataSource
sharding.jdbc.datasource.master0slave0.hikari.driver-class-name=com.mysql.jdbc.Driver
sharding.jdbc.datasource.master0slave0.jdbc-url=jdbc:mysql://192.168.0.3:3306/ds0?characterEncoding=utf-8&serverTimezone=Asia/Shanghai
sharding.jdbc.datasource.master0slave0.username=test
sharding.jdbc.datasource.master0slave0.password=12root
# 从第一个数据库
sharding.jdbc.datasource.master1slave0.type=com.zaxxer.hikari.HikariDataSource
sharding.jdbc.datasource.master1slave0.hikari.driver-class-name=com.mysql.jdbc.Driver
sharding.jdbc.datasource.master1slave0.jdbc-url=jdbc:mysql://192.168.0.3:3306/ds1?characterEncoding=utf-8&serverTimezone=Asia/Shanghai
sharding.jdbc.datasource.master1slave0.username=test
sharding.jdbc.datasource.master1slave0.password=12root

# 读写分离配置
# 从库的读取规则为round_robin(轮询策略),除了轮询策略,还有支持random(随机策略)
sharding.jdbc.config.masterslave.load-balance-algorithm-type=round_robin
# 逻辑主从库名和实际主从库映射关系
# 主数据库0
sharding.jdbc.config.sharding.master-slave-rules.ds0.master-data-source-name=master0
# 从数据库0
sharding.jdbc.config.sharding.master-slave-rules.ds0.slave-data-source-names=master0slave0
# 主数据库1
sharding.jdbc.config.sharding.master-slave-rules.ds1.master-data-source-name=master1
# 从数据库1
sharding.jdbc.config.sharding.master-slave-rules.ds1.slave-data-source-names=master1slave0


# 分库分表配置
# 水平拆分的数据库(表) 配置分库 + 分表策略 行表达式分片策略
# 分库策略
sharding.jdbc.config.sharding.default-database-strategy.inline.sharding-column=id
sharding.jdbc.config.sharding.default-database-strategy.inline.algorithm-expression=ds$->{id % 2}
# 分表策略 其中user为逻辑表 分表主要取决于age行
sharding.jdbc.config.sharding.tables.user.actual-data-nodes=ds$->{0..1}.user_$->{0..1}
sharding.jdbc.config.sharding.tables.user.table-strategy.inline.sharding-column=age
# 分片算法表达式
sharding.jdbc.config.sharding.tables.user.table-strategy.inline.algorithm-expression=user_$->{age % 2}

# 主键 UUID 18位数 如果是分布式还要进行一个设置 防止主键重复
#sharding.jdbc.config.sharding.tables.user.key-generator-column-name=id

# 打印操作的sql以及库表数据等
sharding.jdbc.config.props.sql.show=true
spring.main.allow-bean-definition-overriding=true

其他项目配置不变,和之前保持一致即可

三. 测试

1.查询全部数据库

打开浏览器输入 http://localhost:8080/select

file

file

2.插入数据

打开浏览器 分别访问

http://localhost:8080/insert?id=1&name=lhd&age=12
http://localhost:8080/insert?id=2&name=lhd&age=13
http://localhost:8080/insert?id=3&name=lhd&age=14
http://localhost:8080/insert?id=4&name=lhd&age=15

控制台打印

file

结果和之前的一样
根据分片算法和分片策略,不同的id以及age取模落入不同的库表 达到了分库分表

3.查询全部数据

打开浏览器输入 http://localhost:8080/select

file

控制台打印

file

四. 问题

1. 无法知道走的到底是哪个数据源
相信大家也发现了,当读写分离和分库分表集成时
虽然我们配置sql.show=true
但是控制台最终打印不出所执行的数据源是哪个
不知道是从库还是主库
2.读写分离实现

读写分离的流程

获取主从库配置规则,数据源封装成MasterSlaveDataSource
根据ShardingMasterSlaveRouter路由计算,得到sqlRouteResult.getRouteUnits()单元列表,然后将结果addAll添加并返回
执行每个RouteUnits的时候需要获取连接,这里根据轮询负载均衡算法RoundRobinMasterSlaveLoadBalanceAlgorithm得到从库数据源,拿到连接后就开始执行具体的SQL查询了,这里通过PreparedStatementHandler.execute()得到执行结果
结果归并后返回

MasterSlaveDataSource.class

package io.shardingsphere.shardingjdbc.jdbc.core.datasource;

import io.shardingsphere.api.ConfigMapContext;
import io.shardingsphere.api.config.rule.MasterSlaveRuleConfiguration;
import io.shardingsphere.core.constant.properties.ShardingProperties;
import io.shardingsphere.core.rule.MasterSlaveRule;
import io.shardingsphere.shardingjdbc.jdbc.adapter.AbstractDataSourceAdapter;
import io.shardingsphere.shardingjdbc.jdbc.core.connection.MasterSlaveConnection;
import io.shardingsphere.transaction.api.TransactionTypeHolder;
import java.sql.Connection;
import java.sql.DatabaseMetaData;
import java.sql.SQLException;
import java.util.Map;
import java.util.Properties;
import javax.sql.DataSource;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

public class MasterSlaveDataSource extends AbstractDataSourceAdapter {
    private static final Logger log = LoggerFactory.getLogger(MasterSlaveDataSource.class);
    private final DatabaseMetaData databaseMetaData;
    private final MasterSlaveRule masterSlaveRule;
    private final ShardingProperties shardingProperties;

    public MasterSlaveDataSource(Map<String, DataSource> dataSourceMap, MasterSlaveRuleConfiguration masterSlaveRuleConfig, Map<String, Object> configMap, Properties props) throws SQLException {
        super(dataSourceMap);
        this.databaseMetaData = this.getDatabaseMetaData(dataSourceMap);
        if (!configMap.isEmpty()) {
            ConfigMapContext.getInstance().getConfigMap().putAll(configMap);
        }

        this.masterSlaveRule = new MasterSlaveRule(masterSlaveRuleConfig);
        // 从配置文件获取配置的主从数据源
        this.shardingProperties = new ShardingProperties(null == props ? new Properties() : props);
    }

    // 获取主从配置关系
    public MasterSlaveDataSource(Map<String, DataSource> dataSourceMap, MasterSlaveRule masterSlaveRule, Map<String, Object> configMap, Properties props) throws SQLException {
        super(dataSourceMap);
        this.databaseMetaData = this.getDatabaseMetaData(dataSourceMap);
        if (!configMap.isEmpty()) {
            ConfigMapContext.getInstance().getConfigMap().putAll(configMap);
        }

        this.masterSlaveRule = masterSlaveRule;
        this.shardingProperties = new ShardingProperties(null == props ? new Properties() : props);
    }

    // 获取数据库元数据
    private DatabaseMetaData getDatabaseMetaData(Map<String, DataSource> dataSourceMap) throws SQLException {
        Connection connection = ((DataSource)dataSourceMap.values().iterator().next()).getConnection();
        Throwable var3 = null;

        DatabaseMetaData var4;
        try {
            var4 = connection.getMetaData();
        } catch (Throwable var13) {
            var3 = var13;
            throw var13;
        } finally {
            if (connection != null) {
                if (var3 != null) {
                    try {
                        connection.close();
                    } catch (Throwable var12) {
                        var3.addSuppressed(var12);
                    }
                } else {
                    connection.close();
                }
            }

        }

        return var4;
    }

    public final MasterSlaveConnection getConnection() {
        return new MasterSlaveConnection(this, this.getShardingTransactionalDataSources().getDataSourceMap(), TransactionTypeHolder.get());
    }

    public DatabaseMetaData getDatabaseMetaData() {
        return this.databaseMetaData;
    }

    public MasterSlaveRule getMasterSlaveRule() {
        return this.masterSlaveRule;
    }

    public ShardingProperties getShardingProperties() {
        return this.shardingProperties;
    }
}

配置文件配置的主从规则
MasterSlaveRule.class

package io.shardingsphere.core.rule;

import com.google.common.base.Preconditions;
import io.shardingsphere.api.algorithm.masterslave.MasterSlaveLoadBalanceAlgorithm;
import io.shardingsphere.api.algorithm.masterslave.MasterSlaveLoadBalanceAlgorithmType;
import io.shardingsphere.api.config.rule.MasterSlaveRuleConfiguration;
import java.util.Collection;

public class MasterSlaveRule {
    //名称(这里是ds0和ds1)
    private final String name;
    //主库数据源名称(这里是ds_master_0和ds_master_1)
    private final String masterDataSourceName;
    //所属从库列表,key为从库数据源名称,value是真实的数据源
    private final Collection<String> slaveDataSourceNames;
    //主从库负载均衡算法
    private final MasterSlaveLoadBalanceAlgorithm loadBalanceAlgorithm;
    //主从库路由配置
    private final MasterSlaveRuleConfiguration masterSlaveRuleConfiguration;

轮询负载均衡算算法
RoundRobinMasterSlaveLoadBalanceAlgorithm.class

package io.shardingsphere.api.algorithm.masterslave;

import java.util.List;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.atomic.AtomicInteger;

//轮询负载均衡策略,按照每个从节点访问次数均衡
public final class RoundRobinMasterSlaveLoadBalanceAlgorithm implements MasterSlaveLoadBalanceAlgorithm {
    private static final ConcurrentHashMap<String, AtomicInteger> COUNT_MAP = new ConcurrentHashMap();

    public RoundRobinMasterSlaveLoadBalanceAlgorithm() {
    }

    public String getDataSource(String name, String masterDataSourceName, List<String> slaveDataSourceNames) {
        AtomicInteger count = COUNT_MAP.containsKey(name) ? (AtomicInteger)COUNT_MAP.get(name) : new AtomicInteger(0);
        COUNT_MAP.putIfAbsent(name, count);
        count.compareAndSet(slaveDataSourceNames.size(), 0);
        return (String)slaveDataSourceNames.get(Math.abs(count.getAndIncrement()) % slaveDataSourceNames.size());
    }
}

ShardingMasterSlaveRouter.class

//
// Source code recreated from a .class file by IntelliJ IDEA
// (powered by Fernflower decompiler)
//

package io.shardingsphere.core.routing.router.masterslave;

import io.shardingsphere.core.constant.SQLType;
import io.shardingsphere.core.hint.HintManagerHolder;
import io.shardingsphere.core.routing.RouteUnit;
import io.shardingsphere.core.routing.SQLRouteResult;
import io.shardingsphere.core.rule.MasterSlaveRule;
import java.beans.ConstructorProperties;
import java.util.ArrayList;
import java.util.Collection;
import java.util.Iterator;
import java.util.LinkedList;

public final class ShardingMasterSlaveRouter {
    private final Collection<MasterSlaveRule> masterSlaveRules;

    // 得到最终的sql路由
    public SQLRouteResult route(SQLRouteResult sqlRouteResult) {
        Iterator var2 = this.masterSlaveRules.iterator();

        while(var2.hasNext()) {
            MasterSlaveRule each = (MasterSlaveRule)var2.next();
            this.route(each, sqlRouteResult);
        }

        return sqlRouteResult;
    }

    //进行计算筛选得到最终sql路由
    private void route(MasterSlaveRule masterSlaveRule, SQLRouteResult sqlRouteResult) {
        Collection<RouteUnit> toBeRemoved = new LinkedList();
        Collection<RouteUnit> toBeAdded = new LinkedList();
        Iterator var5 = sqlRouteResult.getRouteUnits().iterator();

        while(var5.hasNext()) {
            RouteUnit each = (RouteUnit)var5.next();
            if (masterSlaveRule.getName().equalsIgnoreCase(each.getDataSourceName())) {
                toBeRemoved.add(each);
                if (this.isMasterRoute(sqlRouteResult.getSqlStatement().getType())) {
                    MasterVisitedManager.setMasterVisited();
                    toBeAdded.add(new RouteUnit(masterSlaveRule.getMasterDataSourceName(), each.getSqlUnit()));
                } else {
                    toBeAdded.add(new RouteUnit(masterSlaveRule.getLoadBalanceAlgorithm().getDataSource(masterSlaveRule.getName(), masterSlaveRule.getMasterDataSourceName(), new ArrayList(masterSlaveRule.getSlaveDataSourceNames())), each.getSqlUnit()));
                }
            }
        }
        //路由移除(查询时 移除所有主库)
        sqlRouteResult.getRouteUnits().removeAll(toBeRemoved);
        //添加从库/主库 具体事件定
        sqlRouteResult.getRouteUnits().addAll(toBeAdded);
    }

    // 判断是不是主库
    private boolean isMasterRoute(SQLType sqlType) {
        return SQLType.DQL != sqlType || MasterVisitedManager.isMasterVisited() || HintManagerHolder.isMasterRouteOnly();
    }

    @ConstructorProperties({"masterSlaveRules"})
    public ShardingMasterSlaveRouter(Collection<MasterSlaveRule> masterSlaveRules) {
        this.masterSlaveRules = masterSlaveRules;
    }
}

注: 判断是不是主库的规则为:

private boolean isMasterRoute(SQLType sqlType) {
        return SQLType.DQL != sqlType || MasterVisitedManager.isMasterVisited() || HintManagerHolder.isMasterRouteOnly();
    }

SQL语言的判断

SQL语言共分为四大类:数据查询语言DQL,数据操纵语言DML,数据定义语言DDL,数据控制语言DCL。

通过断点,查询全部数据时最终的sql路由为

file
走的从库的四个从表
前面的问题也就迎刃而解

目前读写分离和分库分表就完成

源码分析不对,如有错误请指点一二

源码下载: https://github.com/LiHaodong888/SpringBootLearn

本文目录