如何创建基于Hazelcast的DataSource适配Spring Batch

你之前的思路存在核心误区：Hazelcast并非关系型数据库，没有对应的JDBC Driver，无法通过DriverDataSource创建指向它的JDBC数据源。Spring Batch默认的作业元数据存储依赖JDBC，但要改用Hazelcast作为持久层，需要基于Hazelcast的分布式Map（IMap）实现自定义的作业元数据访问层，而非硬套JDBC模式。

核心修正方向 #

1. 移除无用的JDBC DataSource配置，改用Hazelcast IMap存储Spring Batch的作业元数据（JobInstance、JobExecution、StepExecution、ExecutionContext等）
2. 使用Spring Batch提供的MapJobRepositoryFactoryBean生成基于Map的JobRepository，替换内部默认的本地Map为Hazelcast分布式IMap
3. 适配JobExplorer，使用基于Hazelcast IMap的Dao实现

完整配置示例 #

@Configuration
public class HazelcastBatchConfig {

    // 配置Hazelcast客户端实例
    @Bean
    public HazelcastInstance hazelcastInstance() {
        ClientConfig config = new ClientConfig();
        config.setClusterName("dev");
        return HazelcastClient.newHazelcastClient(config);
    }

    // 定义存储作业元数据的Hazelcast分布式Map
    @Bean
    public IMap<Long, JobInstance> jobInstanceMap(HazelcastInstance hazelcastInstance) {
        return hazelcastInstance.getMap("spring-batch-job-instances");
    }

    @Bean
    public IMap<Long, JobExecution> jobExecutionMap(HazelcastInstance hazelcastInstance) {
        return hazelcastInstance.getMap("spring-batch-job-executions");
    }

    @Bean
    public IMap<Long, StepExecution> stepExecutionMap(HazelcastInstance hazelcastInstance) {
        return hazelcastInstance.getMap("spring-batch-step-executions");
    }

    @Bean
    public IMap<Long, ExecutionContext> jobExecutionContextMap(HazelcastInstance hazelcastInstance) {
        return hazelcastInstance.getMap("spring-batch-job-execution-contexts");
    }

    @Bean
    public IMap<Long, ExecutionContext> stepExecutionContextMap(HazelcastInstance hazelcastInstance) {
        return hazelcastInstance.getMap("spring-batch-step-execution-contexts");
    }

    // 配置Hazelcast事务管理器
    @Bean
    public HazelcastTransactionManager hazelcastTransactionManager(HazelcastInstance hazelcastInstance) {
        return new HazelcastTransactionManager(hazelcastInstance);
    }

    // 基于Hazelcast IMap创建JobRepository
    @SneakyThrows
    @Bean
    public JobRepository hazelcastJobRepository(HazelcastTransactionManager transactionManager,
                                                IMap<Long, JobInstance> jobInstanceMap,
                                                IMap<Long, JobExecution> jobExecutionMap,
                                                IMap<Long, StepExecution> stepExecutionMap,
                                                IMap<Long, ExecutionContext> jobExecutionContextMap,
                                                IMap<Long, ExecutionContext> stepExecutionContextMap) {
        MapJobRepositoryFactoryBean factoryBean = new MapJobRepositoryFactoryBean();
        factoryBean.setTransactionManager(transactionManager);
        
        // 替换本地Map为Hazelcast分布式IMap
        factoryBean.setJobInstanceMap(jobInstanceMap);
        factoryBean.setJobExecutionMap(jobExecutionMap);
        factoryBean.setStepExecutionMap(stepExecutionMap);
        factoryBean.setJobExecutionContextMap(jobExecutionContextMap);
        factoryBean.setStepExecutionContextMap(stepExecutionContextMap);
        
        return factoryBean.getObject();
    }

    // 配置JobBuilderFactory
    @Bean
    public JobBuilderFactory jobBuilderFactory(JobRepository hazelcastJobRepository){
        return new JobBuilderFactory(hazelcastJobRepository);
    }

    // 配置StepBuilderFactory
    @Bean
    public StepBuilderFactory stepBuilderFactory(JobRepository hazelcastJobRepository, 
                                                HazelcastTransactionManager transactionManager){
        return new StepBuilderFactory(hazelcastJobRepository, transactionManager);
    }

    // 基于Hazelcast IMap创建JobExplorer
    @Bean
    public JobExplorer jobExplorer(IMap<Long, JobInstance> jobInstanceMap,
                                   IMap<Long, JobExecution> jobExecutionMap,
                                   IMap<Long, StepExecution> stepExecutionMap,
                                   IMap<Long, ExecutionContext> jobExecutionContextMap,
                                   IMap<Long, ExecutionContext> stepExecutionContextMap) {
        // 使用适配Hazelcast的Dao实现
        JobInstanceDao jobInstanceDao = new MapJobInstanceDao(jobInstanceMap);
        JobExecutionDao jobExecutionDao = new MapJobExecutionDao(jobExecutionMap, jobExecutionContextMap);
        StepExecutionDao stepExecutionDao = new MapStepExecutionDao(stepExecutionMap, stepExecutionContextMap);
        ExecutionContextDao executionContextDao = new MapExecutionContextDao(jobExecutionContextMap, stepExecutionContextMap);
        
        return new SimpleJobExplorer(jobInstanceDao, jobExecutionDao, stepExecutionDao, executionContextDao);
    }

    @Bean
    public ListableJobLocator listableJobLocator(){
        return new MapJobRegistry();
    }
}

关键注意事项 #

1. 集群一致性：Hazelcast的IMap会自动在集群内同步作业元数据，适合分布式Spring Batch部署场景
2. 数据持久化：如果需要重启集群后保留作业元数据，需为Hazelcast配置持久化存储（如磁盘、数据库）
3. 事务支持：HazelcastTransactionManager会管理IMap的事务操作，确保作业元数据的一致性
4. 移除冗余配置：完全删除原有的JDBC DataSource相关代码，避免Spring Batch自动启用JDBC模式

内容的提问来源于stack exchange，提问作者Jacob Goldverg