Update reference documentation as Spring Batch's resourceless infrastructure means that it no longer always requires a DataSource

[ah1508]

Spring Boot can create the tables used by Spring batch to store statistics about job and step execution if the spring.batch.jdbc.initialize-schema is set to always

See: https://docs.spring.io/spring-boot/how-to/data-initialization.html#howto.data-initialization.batch

With Spring Batch 6, the default behavior does not store anything, because a ResourcelessJobRepository is register in the context so schema initialization does not make sense.

@EnableJdbcJobRepository enables JDBC storage: a SimpleJobRepository is registered in the context and setting spring.batch.jdbc.initialize-schema to always should trigger tables creation:

@SpringBootApplication
@EnableBatchProcessing // otherwise @EnableJdbcJobRepository has no effect
@EnableJdbcJobRepository
public class EntryPoint {

    @Bean
    Job job(JobRepository jobRepository){
        return new JobBuilder("job", jobRepository)
                .start(new StepBuilder("step", jobRepository)
                        .tasklet((_, _)-> RepeatStatus.FINISHED)
                        .build())
                .build();
    }

    public static void main(String[] args) {
        var ctx = SpringApplication.run(EntryPoint.class, "spring.batch.jdbc.initialize-schema=always");
        System.out.println(ctx.getBean(JobRepository.class));
    }
}

But it does not: tables are not created.

If there is something more to do it should be written in the documentation somewhere. Search for spring.batch.jdbc.initialize-schema in the Spring batch documentation does not give any result related to schema creation (probably because it's Spring boot responsability).

Sample here with the code hereabove and a h2 database and spring-boot-starter-batch-jdbc

demo-issue.zip

[ah1508]

The BatchJdbcAutoConfiguration from spring-boot-batch-jdbc has following annotation

@ConditionalOnMissingBean(value = DefaultBatchConfiguration.class, annotation = EnableBatchProcessing.class)

Since None of the requirements must be met for the condition to match, it explains why the schema creation is ignored. But removing @EnableBatchProcessing makes @EnableJdbcJobRepository useless because "This annotation should be used on a @configuration class annotated with EnableBatchProcessing."

Removing both @EnableBatchProcessing and @EnableJdbcJobRepository solves the problem and Spring Boot finds its way to configure JobRepository properly, however @EnableJdbcJobRepository is useful to use a specific datasource and a specific PlatformTransactionManager for job execution statistics:

@EnableJdbcJobRepository(dataSourceRef = "spring-batch-datasource", transactionManagerRef = "spring-batch-txManager")

[wilkinsona]

This is behaving as described in the documentation. You can either use @EnableJdbcJobRepository or the configuration properties. If you want to retain full auto-configuration and use the configuration properties, please see the documentation on @BatchDataSource and @BatchTransactionManager.

[ah1508]

Hi Andy,

Thanks for you explaination. However:

Spring Boot brings:

• JobOperator bean
• JobRepository bean
• automatic start of job when program is launch (call to main) except if spring.batch.job.enabled is set to false
• schema initialization (with spring-boot-starter-batch-jdbc)

Use of @EnableBatchProcessing disables auto configuration but the JobRepository and JobOperator are provided by BatchRegistrar and the job is still launched. But then you lose the schema initialization, I think it is a missing feature on Spring Batch side or a missing information on the documentation. Or Spring boot could initialize schema even if @EnableBatchProcessing is present.

Until now the only way to keep schema initialization possible is to

• use spring-boot-starter-batch-jdbc,
• not use @EnableBatchProcessing
• use, if needed, @BatchDataSource and @BatchTransactionManager.

The "how to guide" (https://docs.spring.io/spring-boot/reference/io/spring-batch.html) could also be improved:

By default, batch applications require a DataSource to store job details.. Not anymore with Spring Batch 6. That's why I tried the new @EnableJdbcJobRepository, which only works with @EnableBatchProcessing which disables BatchJdbcAutoConfiguration

To take greater control, add @EnableBatchProcessing to one of your @Configuration classes or extend DefaultBatchConfiguration. But the BatchJdbcAutoConfiguration will not apply anymore and you lose spring.batch.jdbc.initialize-schema.

[wilkinsona]

But the JobRepository and JobOperator are provided by BatchRegistrar and the job is still launched

That's standard Spring Batch behavior.

Or Spring boot could initialize schema even if @EnableBatchProcessing is present.

We can't do that. @EnableBatchProcessing is the signal that the user wants to take complete control over the configuration of Spring Batch so its use causes Spring Boot's auto-configuration to back off. This includes initialization of the database schema. You can initialise it manually using the Batch documentation to guide you.

Until now the only way to keep schema initialization possible is to

• use spring-boot-starter-batch-jdbc,
• not use @EnableBatchProcessing
• use, if needed, @BatchDataSource and @BatchTransactionManager.

Correct. @BatchDataSource and @BatchTransactionManager were introduced for exactly this purpose – taking control over the DataSource and or TransactionManager that's used while retaining the rest of Boot's auto-configuration, including schema initialization.

By default, batch applications require a DataSource to store job details.. Not anymore with Spring Batch 6.

We can use this issue to update this.

To take greater control, add @EnableBatchProcessing to one of your @configuration classes or extend DefaultBatchConfiguration. But the BatchJdbcAutoConfiguration will not apply anymore

This has always been the case. By "take greater control" it means "switch off the auto-configuration" and that includes schema initialization. I've opened #48264.