Spring Batch 批处理（7） - 异常处理及容错机制

@Component("restartDemoReader")
public class RestartDemoReader implements ItemStreamReader<Customer> {
 
    // 记录当前读取的行数
    private Long curLine = 0L;
    // 重启状态初始值
    private boolean restart = false;
 
    private FlatFileItemReader<Customer> reader = new FlatFileItemReader<>();
 
    // 持久化信息到数据库
    private ExecutionContext executionContext;
    RestartDemoReader
    public () {
        
        reader.setResource(new ClassPathResource("restartDemo.csv"));
 
        DelimitedLineTokenizer tokenizer = new DelimitedLineTokenizer();
        tokenizer.setNames(new String[]{"id", "firstName", "lastName", "birthdate"});
 
        DefaultLineMapper<Customer> lineMapper = new DefaultLineMapper<>();
        lineMapper.setLineTokenizer(tokenizer);
        lineMapper.setFieldSetMapper((fieldSet -> {
            return Customer.builder().id(fieldSet.readLong("id"))
                    .firstName(fieldSet.readString("firstName"))
                    .lastName(fieldSet.readString("lastName"))
                    .birthdate(fieldSet.readString("birthdate"))
                    .build();
        }));
        lineMapper.afterPropertiesSet();
 
        reader.setLineMapper(lineMapper);
    }
 
    @Override
    public Customer read() throws Exception, UnexpectedInputException, ParseException,
            NonTransientResourceException {
 
        Customer customer = null;
 
        this.curLine++;
        //如果是重启，则从上一步读取的行数继续往下执行
        if (restart) {
            reader.setLinesToSkip(this.curLine.intValue()-1);
            restart = false;
            System.out.println("Start reading from line: " + this.curLine);
        }
 
        reader.open(this.executionContext);
 
        customer = reader.read();
        //当匹配到wrongName时，显示抛出异常，终止程序
        if (customer != null) {
            if (customer.getFirstName().equals("wrongName"))
                throw new RuntimeException("Something wrong. Customer id: " + customer.getId());
        } else {
            curLine--;
        }
        return customer;
    }
 
    /**
     * 判断是否是重启job
     * @param executionContext
     * @throws ItemStreamException
     */
    @Override
    public void open(ExecutionContext executionContext) throws ItemStreamException {
        this.executionContext = executionContext;
				// 如果是重启job，从数据库读取重启的行数，从重启行数开始重新执行
        if (executionContext.containsKey("curLine")) {
            this.curLine = executionContext.getLong("curLine");
            this.restart = true;
        }
				// 如果不是重启job，初始化行数，从第一行开始执行
				else {
            this.curLine = 0L;
            executionContext.put("curLine", this.curLine.intValue());
        }
 
    }
 
    @Override
    public void update(ExecutionContext executionContext) throws ItemStreamException {
		    // 每执行完一个批次chunk，打印当前行数
        System.out.println("update curLine: " + this.curLine);
        executionContext.put("curLine", this.curLine);
 
    }
 
    @Override
    public void close() throws ItemStreamException {
 
    }
}

@Configuration
public class RestartDemoJobConfiguration {
    @Autowired
    private JobBuilderFactory jobBuilderFactory;
 
    @Autowired
    private StepBuilderFactory stepBuilderFactory;
 
    @Autowired
    @Qualifier("flatFileDemoWriter")
    private ItemWriter<? super Customer> flatFileDemoWriter;
 
    @Autowired
    @Qualifier("restartDemoReader")
    private ItemReader<Customer> restartDemoReader;
 
    @Bean
    public Job restartDemoJob(){
        return jobBuilderFactory.get("restartDemoJob")
                .start(restartDemoStep())
                .build();
 
    }
 
    @Bean
    public Step restartDemoStep() {
        return stepBuilderFactory.get("restartDemoStep")
                .<Customer,Customer>chunk(10)
                .reader(restartDemoReader)
                .writer(flatFileDemoWriter)
                .build();
    }
}

<br/>


这里设置了允许重试的次数为3次，允许跳过的数据最多为1条，如果job失败了，运行重跑次数最多为3次。

<br/>
在skip后面配置跳过错误的监听器SkipListener

// 发生写操作跳过错误时，需要执行的监听
public void onSkipInWrite(String item, Throwable t){
}

// 处理数据时跳过错误时，需要执行的监听
public void onSkipInProcess (String item, Throwable t){
	System.out.println(item + "occur exception" + t);
}

<br/>


重新运行程序，可以得到新的结果：

![file](https://imgconvert.csdnimg.cn/aHR0cHM6Ly9ncmFwaC5iYWlkdS5jb20vcmVzb3VyY2UvMjIyNzJjYzk4ZTk5ZTk4Nzg1NWU3MDE1ODM0MjUzNDUucG5n?x-oss-process=image/format,png)

<br/>

这次可以看到，12条数据中总共有11条数据进入到数据库，而过长的008008008008数据，则因为设置了skip，所以容错机制允许它不进入数据库，这次的Spring batch最终没有因为回滚而中断。

<br/>
查阅一下Spring batch的持久化数据表：

![file](https://imgconvert.csdnimg.cn/aHR0cHM6Ly9ncmFwaC5iYWlkdS5jb20vcmVzb3VyY2UvMjIyNjE5YTg3M2VjNGNkNDE0Y2U1MDE1ODM0MjUzNTUucG5n?x-oss-process=image/format,png)

可以看出，的确是有一条数据被跳过了，但因为是我们允许它跳过的，所以整个job顺利完成，即COMPLETED。



参考：

https://blog.csdn.net/chihe9907/article/details/100601523