데이터베이스에서 replication 은 데이터베이스간의 데이터 복제를 의미한다. 하나의 데이터베이스에서 읽기와 쓰기 작업을 동시에 진행할 경우 병목 현상이 발생할 확률이 높다. 이를 방지하기 위해 흔히 Master(Write) 와 Slave(Read) 로 나누어, Master 에서 변경된 데이터를 Slave 로 복제하는 것을 replication 이라고 한다.

Java 어플리케이션에서 분기 처리하는 방법은 커넥션 풀(DataSource) 을 Master 와 Slave 로 나누어 주고, JDBC 의 Connection.setReadOnly 메소드나 Spring 의 @Transactional 를 사용하는 것이다.

Spring Boot 에서의 read / write 분기 처리

1. DB 설정 (application.yml)

spring:
  datasource:
    master:
      hikari:
        driver-class-name: com.mysql.cj.jdbc.Driver
        url: jdbc:mysql://dev.cluster-abc.rds.amazonaws.com:3306/kps
        read-only: false
        username: username
        password: password

    slave:
      hikari:
        driver-class-name: com.mysql.cj.jdbc.Driver
        url: jdbc:mysql://dev.cluster-ro-abc.rds.amazonaws.com:3306/kps
        read-only: true
        username: username
        password: password

2. DataSource Bean 등록

@Configuration
public class DataSourceConfiguration {

    @Bean
    @ConfigurationProperties(prefix = "spring.datasource.master.hikari")
    public DataSource masterDataSource() {
        return DataSourceBuilder.create()
                .type(HikariDataSource.class)
                .build();
    }

    @Bean
    @ConfigurationProperties(prefix = "spring.datasource.slave.hikari")
    public DataSource slaveDataSource() {
        return DataSourceBuilder.create()
                .type(HikariDataSource.class)
                .build();
    }
}

3. AbstractRoutingDataSource 구현

: Read-Write 가 동시에 필요하거나 복제 지연으로의 이슈를 막기 위해 AbstractRoutingDataSource 구현.

public class routingDataSourceImpl extends AbstractRoutingDataSource {
    @Override
    protected Object determineCurrentLookupKey() {
        return (TransactionSynchronizationManager.isCurrentTransactionReadOnly()) ? "slave" : "master";
    }
}

4. Lazy 설정

: LazyConnectionDataSourceProxy 설정으로 실제 쿼리 직전에 connection 을 생성하여 불필요한 커넥션 점유 방지.

@Configuration(proxyBeanMethods = false)
public class DataSourceConfiguration {

    @Primary
    @Bean
    public DataSource dataSource(@Qualifier("routingDataSource") DataSource routingDataSource) {
        return new LazyConnectionDataSourceProxy(routingDataSource);
    }

    @Bean
    public DataSource routingDataSource(@Qualifier("masterDataSource") DataSource masterDataSource,
                                        @Qualifier("slaveDataSource") DataSource slaveDataSource) {
        RoutingDataSourceImpl routingDataSourceImpl = new RoutingDataSourceImpl();
        Map<Object, Object> targetDataSource = new HashMap<>();
        targetDataSource.put("master", masterDataSource);
        targetDataSource.put("slave", slaveDataSource);
        routingDataSourceImpl.setTargetDataSources(targetDataSource);
        routingDataSourceImpl.setDefaultTargetDataSource(masterDataSource);
        return routingDataSourceImpl;
    }

    @Bean
    @ConfigurationProperties(prefix = "spring.datasource.master.hikari")
    public DataSource masterDataSource() {
        return DataSourceBuilder.create()
                .type(HikariDataSource.class)
                .build();
    }

    @Bean
    @ConfigurationProperties(prefix = "spring.datasource.slave.hikari")
    public DataSource slaveDataSource() {
        return DataSourceBuilder.create()
                .type(HikariDataSource.class)
                .build();
    }
}

5. @Transactional 명시

@Service
public class ExampleService {

    ...

    @Transactional(readOnly = true)
    public List<ExampleEntity> getList() {
        return exampleRepository.getList();
    }

    @Transactional  // default : readOnly = false
    public Integer insertExample() {
        return exampleRepository.insertExample();
    }
}

6. 테스트

위의 코드를 예로 들면 updateExample() 를 실행하여 데이터 삽입을 확인하고, 

getList() 를 실행했을 때 slave dataSource 가 구성되며 삽입한 데이터가 출력되면 성공.