使用Pub/Sub模拟器测试死信队列时触发java.lang.NullPointerException

问题现象 #

1. 测试死信队列功能时，消息达到配置的最大重试次数后，未被正确路由到指定的死信主题
2. 模拟器在重试次数耗尽后出现挂起状态，此时尝试从订阅拉取消息会触发NullPointerException，错误日志如下：

[pubsub] java.lang.NullPointerException
[pubsub]        at com.google.cloud.pubsub.testing.v1.MessageBacklog.pull(MessageBacklog.java:302)
[pubsub]        at com.google.cloud.pubsub.testing.v1.OrderedMessageBacklog.pullFromBacklogs(OrderedMessageBacklog.java:323)
[pubsub]        at com.google.cloud.pubsub.testing.v1.OrderedMessageBacklog.pull(OrderedMessageBacklog.java:181)
[pubsub]        at com.google.cloud.pubsub.testing.v1.OrderedMessageBacklog.pull(OrderedMessageBacklog.java:169)
[pubsub]        at com.google.cloud.pubsub.testing.v1.FakePubsubServer.pull(FakePubsubServer.java:1482)
[pubsub]        at com.google.cloud.pubsub.testing.v1.PubsubGrpcServerAdapter$2.pull(PubsubGrpcServerAdapter.java:349)
[pubsub]        at com.google.pubsub.v1.SubscriberGrpc$MethodHandlers.invoke(SubscriberGrpc.java:2037)
[pubsub]        at io.grpc.stub.ServerCalls$UnaryServerCallHandler$UnaryServerCallListener.onHalfClose(ServerCalls.java:182)
[pubsub]        at io.grpc.PartialForwardingServerCallListener.onHalfClose(PartialForwardingServerCallListener.java:35)
[pubsub]        at io.grpc.ForwardingServerCallListener.onHalfClose(ForwardingServerCallListener.java:23)
[pubsub]        at io.grpc.ForwardingServerCallListener$SimpleForwardingServerCallListener.onHalfClose(ForwardingServerCallListener.java:40)
[pubsub]        at io.grpc.Contexts$ContextualizedServerCallListener.onHalfClose(Contexts.java:86)
[pubsub]        at io.grpc.internal.ServerCallImpl$ServerStreamListenerImpl.halfClosed(ServerCallImpl.java:351)
[pubsub]        at io.grpc.internal.ServerImpl$JumpToApplicationThreadServerStreamListener$1HalfClosed.runInContext(ServerImpl.java:860)
[pubsub]        at io.grpc.internal.ContextRunnable.run(ContextRunnable.java:37)
[pubsub]        at io.grpc.internal.SerializingExecutor.run(SerializingExecutor.java:133)
[pubsub]        at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)
[pubsub]        at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)
[pubsub]        at java.base/java.lang.Thread.run(Thread.java:829)
[pubsub] 
[pubsub] Mar 01, 2024 6:08:18 PM io.gapi.emulators.netty.HttpHandler$1 onError

解决方案 #

1. 升级Pub/Sub模拟器到最新版本 #

旧版本模拟器存在死信队列处理逻辑的Bug，尤其是重试次数耗尽后的消息路由和拉取逻辑。执行以下命令升级：

gcloud components update pubsub-emulator

2. 校验死信队列配置 #

确保订阅的死信配置完全正确：

• 提前创建死信主题，且订阅配置的死信主题名称与实际创建的一致
• 确认maxDeliveryAttempts（最大重试次数）设置为大于0的有效值，未被代码或命令行参数错误覆盖
• 订阅的ackDeadlineSeconds（确认超时）配置与重试策略匹配，避免因超时导致的异常重试计数

3. 规避有序消息相关Bug #

从错误栈可以看出问题出在有序消息的Backlog处理逻辑中，若业务场景不需要有序消息，创建订阅时关闭该特性：

• 使用gcloud命令创建订阅时添加参数：--enable-message-ordering=false
• 代码中创建订阅时设置enableMessageOrdering为false

4. 增加拉取延迟 #

模拟器处理死信消息路由存在异步延迟，重试次数耗尽后不要立即拉取原订阅，等待3-5秒后再操作，或直接监听死信主题的订阅来验证消息是否已路由。

5. 临时手动路由（Workaround） #

若上述方法无效，可在重试次数耗尽后，手动将消息发布到死信主题，同时记录该问题到Google的官方Issue跟踪平台。

内容的提问来源于stack exchange，提问作者ppeko