当从EventBus@ConsumeEvent调用时，Quarkus Panache存储库调用以静默方式失败

Question

Quarkus 1.8.3。最终的

直接调用访问PanacherRepository的方法可以按预期工作，但是当通过EventBus调用同一方法时，调用到达该方法并执行每一行，直到它到达任何存储库调用，然后在没有任何发生的指示的情况下无声地失败/退出。

根据日志，直接调用在Quarkus主线程中执行，事件总线调用在vert中执行。x-eventloop-thread-2。

还尝试了以下步骤的组合，结果相同：

• 将EventBus的消费者端包装到一个叛变的Uni

@ApplicationScoped
public class TestEntityRepository implements PanacheRepository<TestEntity> { }

@Slf4j
@Startup
@Transactional
@ApplicationScoped
public class TestService {

    @Inject
    TestEntityRepository repository;

    @Inject
    EventBus eventBus;

    public void startUp(@Observes StartupEvent event) {
        // Call via the event bus prints the log in listEntities() and silently stops when it reaches the repository call.
        eventBus.sendAndForget("test_topic", "eventbus call");
        // This  prints the log in listEntities() and then lists all the entities in the repository.
        listEntities("direct call");
    }

    @ConsumeEvent("test_topic")
    public void listEntities(String testMessage) {
        log.info("Printing all entities via: " + testMessage);
        repository.findAll().stream().map(TestEntity::toString).forEach(log::info);
    }

以下是日志中EventBus部分的摘录：

2020-10-19 21:24:39,612 INFO  [org.acm.com.TestService] (vert.x-eventloop-thread-1) Printing all entities via: eventbus call
2020-10-19 21:24:39,617 TRACE [com.arj.ats.jta] (vert.x-eventloop-thread-1) TransactionImple.setRollbackOnly
2020-10-19 21:24:39,617 TRACE [com.arj.ats.arjuna] (vert.x-eventloop-thread-1) BasicAction::preventCommit( BasicAction: 0:ffff7f000101:a721:5f8de7f7:0 status: ActionStatus.RUNNING)
2020-10-19 21:24:39,617 TRACE [com.arj.ats.jta] (vert.x-eventloop-thread-1) TransactionImple.getStatus: javax.transaction.Status.STATUS_MARKED_ROLLBACK
2020-10-19 21:24:39,618 TRACE [com.arj.ats.jta] (vert.x-eventloop-thread-1) BaseTransaction.rollback
2020-10-19 21:24:39,618 TRACE [com.arj.ats.jta] (vert.x-eventloop-thread-1) TransactionImple.rollbackAndDisassociate
2020-10-19 21:24:39,618 TRACE [com.arj.ats.arjuna] (vert.x-eventloop-thread-1) BasicAction::Abort() for action-id 0:ffff7f000101:a721:5f8de7f7:0
2020-10-19 21:24:39,618 TRACE [com.arj.ats.arjuna] (vert.x-eventloop-thread-1) BasicAction::removeChildThread () action 0:ffff7f000101:a721:5f8de7f7:0 removing 1
2020-10-19 21:24:39,618 TRACE [com.arj.ats.arjuna] (vert.x-eventloop-thread-1) BasicAction::removeChildThread () action 0:ffff7f000101:a721:5f8de7f7:0 removing 1 result = true
2020-10-19 21:24:39,618 TRACE [com.arj.ats.arjuna] (vert.x-eventloop-thread-1) TransactionReaper::remove ( BasicAction: 0:ffff7f000101:a721:5f8de7f7:0 status: ActionStatus.ABORTED )

更新：bugreport已创建

Answer 1

事实证明，EventBus正在吞咽关于阻塞调用的异常，并且返回一个像Uni一样的反应响应，甚至显式地将数据库调用包装到Uni中，仍将使用相同的线程，而不是预期的工作线程。

解决方案是在事件侦听器上使用blocking=true。因此，它没有设置预期的行为（如我所想），而是为阻塞调用准备事件循环。。

工作守则是：

@Slf4j
@Startup
@ApplicationScoped
public class TestService {

    @Inject
    TestEntityRepository repository;

    @Inject
    EventBus eventBus;

    public void startUp(@Observes StartupEvent event) {
        eventBus.sendAndForget("test_topic", "eventbus call");
    }

    @ConsumeEvent(value = "test_topic", blocking = true)
    @Transactional
    public Uni<Void> listEntities(String testMessage) {
        log.info("Printing all entities via: " + testMessage);
        try {
            repository.findAll().forEach(log::info);
            return Uni.createFrom().voidItem();
        } catch (Exception e) {
            return Uni.createFrom().failure(e);
        }
    }
}