Spring Batch Core Concepts: Job, Step, Chunk, and Tasklet

Q: Spring Batch Core Concepts: Job, Step, Chunk, and Tasklet

A **Job** is the top-level container representing an entire batch process, composed of one or more **Steps** executed in sequence or conditionally. Each Step follows either a **chunk-oriented model** (read-process-write in fixed-size chunks inside one transaction) or a **Tasklet model** (a single custom unit of work). The **JobRepository** persists every JobInstance, JobExecution, and StepExecution to a relational database, which is what enables restartability, audit trails, and skip/retry tracking — features that distinguish Spring Batch from a hand-rolled batch loop.

Intermediate·2 min read·

AmazonGoldman SachsJP MorganCapgeminiInfosys

Fast recall

Quick revision

Core answer

A Job is the top-level container representing an entire batch process, composed of one or more Steps executed in sequence or conditionally. Each Step follows either a chunk-oriented model (read-process-write in fixed-size chunks inside one transaction) or a Tasklet model (a single custom unit of work). The JobRepository persists every JobInstance, JobExecution, and StepExecution to a relational database, which is what enables restartability, audit trails, and skip/retry tracking — features that distinguish Spring Batch from a hand-rolled batch loop.

Interview-ready guide

Interview answer

Spring Batch structures batch processing around a clear hierarchy: Job, Step, and either Chunk or Tasklet at the step level. Every layer is persisted to the JobRepository, which is what makes restart and audit possible.

The hierarchy

Job — the overall batch process, like 'import monthly invoices'; composed of one or more Steps in sequence or conditionally
Step — one phase of the job; each Step has its own execution status and can be restarted independently
Chunk — the read-process-write model for a step; reads N items, processes each, writes the batch — all inside one transaction
Tasklet — a single operation for a step; not a read-process-write loop; execute() returns RepeatStatus.FINISHED

JobRepository — the persistence layer

Stores JobInstance (logical job run identified by Job name + JobParameters)
Stores JobExecution (one physical attempt at the JobInstance)
Stores StepExecution (per-step state plus serialized ExecutionContext for restart)
Spring Boot auto-configures it from the application's DataSource

Chunk transaction model

Chunk size 100 means: read 100 records, process each, write all 100 in one transaction
A write failure rolls back only that chunk — already committed chunks are not re-processed on restart
Larger chunks: fewer transactions, better throughput; smaller chunks: finer-grained checkpointing

Tasklet vs Chunk

Chunk model is for streaming work — rows in a table, lines in a file
Tasklet is for one-shot work — truncating a table, sending a notification, calling a stored procedure
They are mutually exclusive within a single Step

In an interview, I'd emphasize the JobRepository — most candidates skip it, but it's what makes Spring Batch suitable for production rather than a hand-rolled loop.

03 · Deep dive