lulu lulu - 7 months ago 28
Java Question

Spring-batch flow / split after a step

I am building a spring-batch solution that contains the following process:

step 1 : split a list into multiple lists
step 2 : process each sub-list
step 3 : merge sub-lists

The generated sub-lists can be processed in parallel, and according to the spring-batch documentation this is supported. Sadly I can only find spring-batch example jobs that start with parallel steps, not examples that start out sequentially.

The following job will not compile. Spring gives me an error: 'cannot resolve step2'

<batch:job id="webServiceJob2">
<batch:step id="step1" next="step2"></batch:step>
<batch:split id="step2" next="step3"></batch:split>
<batch:step id="step3"></batch:step>
</batch:job>


So how can I configure a job to first run a single step, than run a number of steps in parallel, and then run a last single step?

Answer

I've stumbled upon this question asking about how split works, and maybe this answer arrives a bit (one year) late, but here I go...

The issue there is "split" is not a step by itself, but you were naming (and referencing) it as it was:

<batch:job id="webServiceJob2">
    <batch:step id="step1" next="step2"></batch:step>
    <batch:split id="step2" next="step3"></batch:split> <!-- This is not a step -->
    <batch:step id="step3"></batch:step>
</batch:job>

The correct syntax would be:

<batch:job id="webServiceJob2">
    <batch:step id="step1" next="step2"></batch:step>
    <batch:split id="split_step2" next="step3">
        <flow> 
             <step id="step2_A_1" ... next="step2_A_2"/>
             <step id="step2_A_2" ... />
        </flow>
        <flow> 
             <step id="step2_B_1" ... />
        </flow>
    </batch:split>
    <batch:step id="step3"></batch:step>
</batch:job>

But this is not what you want to achieve, because by split declarations you have to set in compile time the exact number of parallel steps that will be executed, and the purpose of split is using different steps in each flow instead calling several times the same one.

You should check the documentation about Scaling and Parallel processes, the partition step seems a good candidate for your requirements.