Spring Scheduler/Executor 的 AOP 特性

Question

I have been working on an old project where both Spring scheduler is enabled (@Scheduled actively being used) also some native JDK thread pool instances active too. In the project configuration xml I see below;

<task:scheduler id="taskScheduler" pool-size="${task-scheduler.pool-size}"/>
<task:executor id="taskExecutor" pool-size="${task-executor.pool-size}" queue-capacity="${task-executor.queue-capacity}"/>
<task:annotation-driven executor="taskExecutor" scheduler="taskScheduler"/>

And them some quartz implementation comes arise with its own job definition, trigger definition stuff where the Jobs defines their own ThreadPoolExecutors as below,

BlockingQueue<Runnable> workerTaskQueue = new ArrayBlockingQueue<Runnable>(poolSize*3);
 threadPoolExecutor = new 
     ThreadPoolExecutor(poolSize,poolSize,1000L,TimeUnit.MILLISECONDS,workerTaskQueue);

then starts to submit tasks (Runnables) into the pool.

threadPoolExecutor.execute(new ImcpWorker(task, this, workerTaskUtil));

But what I see is that at some point the Spring task rejection exception thrown for these tasks. This is nonsense (unless Spring intercepts using AOP the thread pool executors, even if they are created natively). Because there is no spring managed executor.

2021-06-21 11:51:58,679 ERROR [pool-151-thread-81] LisJobHandler - Exception occured: Executor [java.util.concurrent.ThreadPoolExecutor@5532827b[Running, pool size = 1000, active threads = 1000, queued tasks = 100000, completed tasks = 135592411]] did not accept task: org.springframework.aop.interceptor.AsyncExecutionInterceptor$1@5a237108 msisdn:5363443640 org.springframework.core.task.TaskRejectedException: Executor [java.util.concurrent.ThreadPoolExecutor@5532827b[Running, pool size = 1000, active threads = 1000, queued tasks = 100000, completed tasks = 135592411]] did not accept task: org.springframework.aop.interceptor.AsyncExecutionInterceptor$1@5a237108

So again the question, does spring scheduler and executors (if configured) intercepts ThreadPoolExecutors in an application ?

Answer 1

Well the issue was out of nowhere about my initial assumption. As I go deeper in spring debugging I see that the queue task submission comes from one of my other bean. It has async task registrations and each error in the app calls this async method to trigger some custom actions. So that once an endpoing failed really bad and can not recover, this issue occurs. Because the current design keeps calling the async method and each call occupies a slot in the executor's pool.