我有一些消息应用程序需要至少一次保证。我从文档中了解到:akka--持久性--是关于行动者状态的。它使用一些逐层的意识形态来处理这种状态。
在文档中,我发现了一些AbstractPersistentActorWithAtLeastOnceDelivery,它似乎提出了这种保证。但是现在我对一般的概念有一些怀疑:
我的演员是一个简单的传输器,我所需要的只是交付保证。所以我实际上并不关心演员的状态和所有消耗宝贵内存的分层。大量的日志会是问题的原因吗?
演员:
public class SafeSenderActor extends AbstractPersistentActorWithAtLeastOnceDelivery {
private String persistenceId;
private ActorSelection destination;
public SafeSenderActor() {
System.out.println("SafeSenderActor created");
this.persistenceId = "safe-persistent-actor-id-" + UUID.randomUUID();
destination = context().actorSelection("/user/safeReceiverRouter");
}
@Override
public String persistenceId() {
return persistenceId;
}
@Override
public AbstractActor.Receive createReceive() {
return ReceiveBuilder.create()
.match(SenderTaskMessage.class, msg -> {
persistAsync(new MsgSentEvent(msg.getTestMessage()), this::updateState);
})
.match(ConfirmRobustMessageDelivery.class, ack -> {
persistAsync(new MsgConfirmEvent(ack.getMessageId(), ack.getLocalMessageNumber()), this::updateState);
})
.build();
}
@Override
public Receive createReceiveRecover() {
return receiveBuilder().match(Object.class, this::updateState).build();
}
private void updateState(Object event) {
if (event instanceof MsgSentEvent) {
MsgSentEvent ev = (MsgSentEvent) event;
deliver(destination, deliveryId -> new RobustTestMessage(deliveryId, ev.getMessage()));
} else if (event instanceof MsgConfirmEvent) {
MsgConfirmEvent ev = (MsgConfirmEvent) event;
confirmDelivery(ev.getDeliveryId());
}
}
}
在一些连续的消息之后,我收到错误:
[ERROR] [07/06/2017 01:40:33.446] [sender-system-akka.actor.default-dispatcher-50] [akka://sender-system@127.0.0.1:6666/user/safeSendersRouter/$d] Failed to persist event type [com.test.common.events.MsgSentEvent] with sequence number [358698] for persistenceId [safe-persistent-actor-id-648ec66d-7b7f-4291-b3a2-9bd395d92dc7]. (akka.pattern.CircuitBreaker$$anon$1: Circuit Breaker Timed out.)
我用leveldb作为日志。
1条答案
按热度按时间ecbunoof1#
这可能是一个很长的时间,但也许它仍然是有用的回答。您可以覆盖方法
onPersistFailure
和onPersistRejected
,以调查您提到的错误。onPersistFailure
方法被调用,如果持久失败。演员将停止。最好的做法是在一段时间后再次启动演员,并使用回退管理员。如果JOURNAl无法保存事件,则调用onPersistRejected
方法。参与者为RESUMED。