Master Spring @Async: Custom Thread Pools and Real-World Usage

Overview of @Async

Since Spring 3, the @Async annotation can be placed on a method to execute it asynchronously. The caller returns immediately while the actual execution is submitted to a Spring TaskExecutor, typically a thread pool.

When to Use Custom Thread Pools

In production projects it is recommended to replace the default executor with a custom thread pool by implementing AsyncConfigurer (or extending AsyncConfigurerSupport) and providing a TaskExecutor bean.

Built-in Executors Provided by Spring

SimpleAsyncTaskExecutor – creates a new thread for each task, not a true pool.

SyncTaskExecutor – executes tasks synchronously, no async behavior.

ConcurrentTaskExecutor – an adapter for java.util.concurrent.Executor, rarely needed.

SimpleThreadPoolTaskExecutor – used by Quartz and non-Quartz code.

ThreadPoolTaskExecutor – the most common choice, a wrapper around java.util.concurrent.ThreadPoolExecutor.

Typical Async Method Signatures

void return type for fire-and-forget calls.

Methods with parameters that are passed to the async task.

Methods returning java.util.concurrent.Future (or Spring’s ListenableFuture).

Enabling @Async and Default Executor

When @Async is used without specifying an executor name, Spring falls back to SimpleAsyncTaskExecutor. This executor creates a new thread per task and does not reuse threads, which can lead to high memory consumption.

Void-returning Calls

Annotate the method with @Async. If you need to propagate exceptions, you must catch and re-throw them manually.

Future-returning Calls

CompletableFuture Calls

CompletableFuture does not require @Async; it can be used with any executor to run asynchronous tasks and compose stages.

CompletionStage represents a stage in an asynchronous computation; when one stage finishes it can trigger another.

stage.thenApply(x -> square(x))
     .thenAccept(x -> System.out.print(x))
     .thenRun(() -> System.out.println());

Drawbacks of the Default SimpleAsyncTaskExecutor

Alibaba Java development guidelines advise against using Executors factory methods because they may create unbounded thread pools or queues, leading to OutOfMemoryError. SimpleAsyncTaskExecutor creates a new thread per task unless the concurrencyLimit property is set, which is disabled by default.

Configuring a Custom Thread Pool

Three common approaches:

Implement AsyncConfigurer and override public Executor getAsyncExecutor() to return a ThreadPoolTaskExecutor.

Extend AsyncConfigurerSupport and provide the executor bean.

Define a bean named taskExecutor (or the name referenced by @Async) of type ThreadPoolTaskExecutor.

Example Implementation

Executor.class: ThreadPoolExecutorAdapter -> ThreadPoolExecutor -> AbstractExecutorService -> ExecutorService -> Executor

TaskExecutor.class: ThreadPoolTaskExecutor -> SchedulingTaskExecutor -> AsyncTaskExecutor -> TaskExecutor