分库分表场景

关系型数据库本身比较容易成为系统瓶颈,单机存储容量、连接数、处理能力都有限。当单表的数据量达到1000W或100G以后,由于查询维度较多,即使添加从库、优化索引,做很多操作时性能仍下降严重。此时就要考虑对其进行切分了,切分的目的就在于减少数据库的负担,缩短查询时间。

分库分表用于应对当前互联网常见的两个场景——大数据量和高并发。通常分为垂直拆分和水平拆分两种。

垂直拆分是根据业务将一个库(表)拆分为多个库(表)。如:将经常和不常访问的字段拆分至不同的库或表中。由于与业务关系密切,目前的分库分表产品均使用水平拆分方式。

水平拆分则是根据分片算法将一个库(表)拆分为多个库(表)。如:按照ID的最后一位以3取余,尾数是1的放入第1个库(表),尾数是2的放入第2个库(表)等。

单纯的分表虽然可以解决数据量过大导致检索变慢的问题,但无法解决过多并发请求访问同一个库,导致数据库响应变慢的问题。所以通常水平拆分都至少要采用分库的方式,用于一并解决大数据量和高并发的问题。这也是部分开源的分片数据库中间件只支持分库的原因。

但分表也有不可替代的适用场景。最常见的分表需求是事务问题。同在一个库则不需考虑分布式事务,善于使用同库不同表可有效避免分布式事务带来的麻烦。目前强一致性的分布式事务由于性能问题,导致使用起来并不一定比不分库分表快。目前采用最终一致性的柔性事务居多。分表的另一个存在的理由是,过多的数据库实例不利于运维管理。综上所述,最佳实践是合理地配合使用分库+分表。

Sharding-JDBC简介

Sharding-JDBC是当当应用框架ddframe中,从关系型数据库模块dd-rdb中分离出来的数据库水平分片框架,实现透明化数据库分库分表访问。Sharding-JDBC是继dubbox和elastic-job之后,ddframe系列开源的第3个项目。

定位为轻量级Java框架,在Java的JDBC层提供的额外服务。 它使用客户端直连数据库,以jar包形式提供服务,无需额外部署和依赖,可理解为增强版的JDBC驱动,完全兼容JDBC和各种ORM框架。

  • 适用于任何基于Java的ORM框架,如:JPA, Hibernate, Mybatis, Spring JDBC Template或直接使用JDBC。
  • 基于任何第三方的数据库连接池,如:DBCP, C3P0, BoneCP, Druid, HikariCP等。
  • 支持任意实现JDBC规范的数据库。目前支持MySQL,Oracle,SQLServer和PostgreSQL。

Sharding-JDBC分片策略灵活,可支持等号、between、in等多维度分片,也可支持多分片键。

SQL解析功能完善,支持聚合、分组、排序、limit、or等查询,并支持Binding Table以及笛卡尔积表查询。

 

项目实践

数据准备

准备两个数据库。并在两个库中建好表, 建表sql如下:



DROP TABLE IF EXISTS `user_auth_0`;
CREATE TABLE `user_auth_0` (
  `user_id` bigint(20) NOT NULL,
  `add_date` datetime NOT NULL DEFAULT CURRENT_TIMESTAMP,
  `email` varchar(16) DEFAULT NULL,
  `password` varchar(255) DEFAULT NULL,
  `phone` varchar(16) DEFAULT NULL,
  `remark` varchar(16) DEFAULT NULL,
  PRIMARY KEY (`user_id`),
  UNIQUE KEY `USER_AUTH_PHONE` (`phone`),
  UNIQUE KEY `USER_AUTH_EMAIL` (`email`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;


DROP TABLE IF EXISTS `user_auth_1`;
CREATE TABLE `user_auth_1` (
  `user_id` bigint(20) NOT NULL,
  `add_date` datetime NOT NULL DEFAULT CURRENT_TIMESTAMP,
  `email` varchar(16) DEFAULT NULL,
  `password` varchar(255) DEFAULT NULL,
  `phone` varchar(16) DEFAULT NULL,
  `remark` varchar(16) DEFAULT NULL,
  PRIMARY KEY (`user_id`),
  UNIQUE KEY `USER_AUTH_PHONE` (`phone`),
  UNIQUE KEY `USER_AUTH_EMAIL` (`email`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;



POM配置



<dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter</artifactId>
        </dependency>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-devtools</artifactId>
            <scope>runtime</scope>
        </dependency>
        <dependency>
            <groupId>org.projectlombok</groupId>
            <artifactId>lombok</artifactId>
            <optional>true</optional>
        </dependency>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-test</artifactId>
            <scope>test</scope>
        </dependency>

        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-web</artifactId>
        </dependency>

       <!-- 引入jpa-->
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-data-jpa</artifactId>
        </dependency>
       <!-- 引入mysql-->
        <dependency>
            <groupId>mysql</groupId>
            <artifactId>mysql-connector-java</artifactId>
        </dependency>
        <!-- druid -->
        <dependency>
            <groupId>com.alibaba</groupId>
            <artifactId>druid</artifactId>
            <version>1.1.9</version>
        </dependency>
        <!-- sharding-jdbc -->
        <dependency>
            <groupId>com.dangdang</groupId>
            <artifactId>sharding-jdbc-core</artifactId>
            <version>1.5.4</version>
        </dependency>
        <!-- fastjson -->
        <dependency>
            <groupId>com.alibaba</groupId>
            <artifactId>fastjson</artifactId>
            <version>1.2.51</version>
        </dependency>



application.yml配置



spring:
  jpa:
    properties:
      hibernate:
        dialect: org.hibernate.dialect.MySQL5InnoDBDialect
    show-sql: true
database0:
  driverClassName: com.mysql.jdbc.Driver
  url: jdbc:mysql://localhost:3306/mazhq?serverTimezone=UTC&useUnicode=true&characterEncoding=utf-8
  username: root
  password: 123456
  databaseName: mazhq

database1:
  driverClassName: com.mysql.jdbc.Driver
  url: jdbc:mysql://localhost:3306/liugh?serverTimezone=UTC&useUnicode=true&characterEncoding=utf-8
  username: root
  password: 123456
  databaseName: liugh



分库分表最主要有几个配置

 1. 有多少个数据源 (2个:database0和database1)



@Data
@ConfigurationProperties(prefix = "database0")
@Component
public class Database0Config {
    private String url;
    private String username;
    private String password;
    private String driverClassName;
    private String databaseName;

    public DataSource createDataSource() {
        DruidDataSource result = new DruidDataSource();
        result.setDriverClassName(getDriverClassName());
        result.setUrl(getUrl());
        result.setUsername(getUsername());
        result.setPassword(getPassword());
        return result;
    }
}



2. 用什么列进行分库以及分库算法 (一般是用具体值对2取余判断入哪个库,我采用的是判断值是否大于20)



@Component
public class DatabaseShardingAlgorithm implements SingleKeyDatabaseShardingAlgorithm<Long> {
    @Autowired
    private Database0Config database0Config;
    @Autowired
    private Database1Config database1Config;
    @Override
    public String doEqualSharding(Collection<String> collection, ShardingValue<Long> shardingValue) {
        Long value = shardingValue.getValue();
        if (value <= 20L) {
            return database0Config.getDatabaseName();
        } else {
            return database1Config.getDatabaseName();
        }
    }

    @Override
    public Collection<String> doInSharding(Collection<String> availableTargetNames, ShardingValue<Long> shardingValue) {
        Collection<String> result = new LinkedHashSet<>(availableTargetNames.size());
        for (Long value : shardingValue.getValues()) {
            if (value <= 20L) {
                result.add(database0Config.getDatabaseName());
            } else {
                result.add(database1Config.getDatabaseName());
            }
        }
        return result;
    }

    @Override
    public Collection<String> doBetweenSharding(Collection<String> availableTargetNames, ShardingValue<Long> shardingValue) {
        Collection<String> result = new LinkedHashSet<>(availableTargetNames.size());
        Range<Long> range = shardingValue.getValueRange();
        for (Long value = range.lowerEndpoint(); value <= range.upperEndpoint(); value++) {
            if (value <= 20L) {
                result.add(database0Config.getDatabaseName());
            } else {
                result.add(database1Config.getDatabaseName());
            }
        }
        return result;
    }
}



3. 用什么列进行分表以及分表算法



@Component
public class TableShardingAlgorithm implements SingleKeyTableShardingAlgorithm<Long> {
    @Override
    public String doEqualSharding(Collection<String> tableNames, ShardingValue<Long> shardingValue) {
        for (String each : tableNames) {
            if (each.endsWith(shardingValue.getValue() % 2 + "")) {
                return each;
            }
        }
        throw new IllegalArgumentException();
    }

    @Override
    public Collection<String> doInSharding(Collection<String> tableNames, ShardingValue<Long> shardingValue) {
        Collection<String> result = new LinkedHashSet<>(tableNames.size());
        for (Long value : shardingValue.getValues()) {
            for (String tableName : tableNames) {
                if (tableName.endsWith(value % 2 + "")) {
                    result.add(tableName);
                }
            }
        }
        return result;
    }

    @Override
    public Collection<String> doBetweenSharding(Collection<String> tableNames, ShardingValue<Long> shardingValue) {
        Collection<String> result = new LinkedHashSet<>(tableNames.size());
        Range<Long> range = shardingValue.getValueRange();
        for (Long i = range.lowerEndpoint(); i <= range.upperEndpoint(); i++) {
            for (String each : tableNames) {
                if (each.endsWith(i % 2 + "")) {
                    result.add(each);
                }
            }
        }
        return result;
    }
}



4. 每张表的逻辑表名和所有物理表名和集成调用



@Configuration
public class DataSourceConfig {
    @Autowired
    private Database0Config database0Config;

    @Autowired
    private Database1Config database1Config;

    @Autowired
    private DatabaseShardingAlgorithm databaseShardingAlgorithm;

    @Autowired
    private TableShardingAlgorithm tableShardingAlgorithm;

    @Bean
    public DataSource getDataSource() throws SQLException {
        return buildDataSource();
    }

    private DataSource buildDataSource() throws SQLException {
        //分库设置
        Map<String, DataSource> dataSourceMap = new HashMap<>(2);
        //添加两个数据库database0和database1
        dataSourceMap.put(database0Config.getDatabaseName(), database0Config.createDataSource());
        dataSourceMap.put(database1Config.getDatabaseName(), database1Config.createDataSource());
        //设置默认数据库
        DataSourceRule dataSourceRule = new DataSourceRule(dataSourceMap, database0Config.getDatabaseName());

        //分表设置,大致思想就是将查询虚拟表Goods根据一定规则映射到真实表中去
        TableRule orderTableRule = TableRule.builder("user_auth")
                .actualTables(Arrays.asList("user_auth_0", "user_auth_1"))
                .dataSourceRule(dataSourceRule)
                .build();

        //分库分表策略
        ShardingRule shardingRule = ShardingRule.builder()
                .dataSourceRule(dataSourceRule)
                .tableRules(Arrays.asList(orderTableRule))
                .databaseShardingStrategy(new DatabaseShardingStrategy("user_id", databaseShardingAlgorithm))
                .tableShardingStrategy(new TableShardingStrategy("user_id", tableShardingAlgorithm)).build();
        DataSource dataSource = ShardingDataSourceFactory.createDataSource(shardingRule);
        return dataSource;
    }


    @Bean
    public KeyGenerator keyGenerator() {
        return new DefaultKeyGenerator();
    }



接口测试代码

1、实体类



/**
 * @author mazhq
 * @date 2019/7/30 16:41
 */
@Entity
@Data
@Table(name = "USER_AUTH", uniqueConstraints = {@UniqueConstraint(name = "USER_AUTH_PHONE", columnNames = {"PHONE"}),
@UniqueConstraint(name = "USER_AUTH_EMAIL", columnNames = {"EMAIL"})})
public class UserAuthEntity implements Serializable {
    private static final long serialVersionUID = 7230052310725727465L;
    @Id
    private Long userId;
    @Column(name = "PHONE", length = 16)
    private String phone;
    @Column(name = "EMAIL", length = 16)
    private String email;
    private String password;
    @Column(name = "REMARK",length = 16)
    private String remark;
    @Column(name = "ADD_DATE", nullable = false, columnDefinition = "datetime default now()")
    private Date addDate;
}



 2. Dao层



@Repository
public interface UserAuthDao extends JpaRepository<UserAuthEntity, Long> {
}



 3. controller层



/**
 * @author mazhq
 * @Title: UserAuthController
 * @date 2019/8/1 17:18
 */
@RestController
@RequestMapping("/user")
public class UserAuthController {
    @Autowired
    private UserAuthDao userAuthDao;

    @PostMapping("/save")
    public String save(){
        for (int i=0;i<40;i++) {
            UserAuthEntity userAuthEntity = new UserAuthEntity();
            userAuthEntity.setUserId((long)i);
            userAuthEntity.setAddDate(new Date());
            userAuthEntity.setEmail("test"+i+"@163.com");
            userAuthEntity.setPassword("123456");
            userAuthEntity.setPhone("1388888888"+i);
            Random r = new Random();
            userAuthEntity.setRemark(""+r.nextInt(100));
            userAuthDao.save(userAuthEntity);
        }
        return "success";
    }

    @PostMapping("/select")
    public String select(){
        return JSONObject.toJSONString(userAuthDao.findAll(Sort.by(Sort.Order.desc("remark"))));
    }
}



测试方式:

先调用:http://localhost:8080/user/save

再查询:http://localhost:8080/user/select