在调用堆栈外部调用 OperationFuture 时 SynchronizationContext 出现异常

Question

我正在使用 Google 的 DocumentAI SDK，但这个错误似乎源于 gRPC SDK。我在 DocumentAI 中调用异步操作，其中 returns 一个 OperationFuture。当我在创建未来的调用堆栈框架内调用方法 OperationFuture.get() 时，代码会正确阻塞，直到未来完成并正常继续。但是，如果创建未来的方法 returns 并且我在其创建框架之外调用 OperationFuture.get() 我总是会遇到以下堆栈跟踪的异常

io.grpc.internal.ManagedChannelImpl uncaughtException
SEVERE: [Channel<1>: (us-documentai.googleapis.com:443)] Uncaught exception in the SynchronizationContext. Panic!
java.util.concurrent.RejectedExecutionException: Task java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask@60d4b478 rejected from java.util.concurrent.ScheduledThreadPoolExecutor@1e3a60f5[Terminated, pool size = 0, active threads = 0, queued tasks = 0, completed tasks = 0]
    at java.util.concurrent.ThreadPoolExecutor$AbortPolicy.rejectedExecution(ThreadPoolExecutor.java:2063)
    at java.util.concurrent.ThreadPoolExecutor.reject(ThreadPoolExecutor.java:830)
    at java.util.concurrent.ScheduledThreadPoolExecutor.delayedExecute(ScheduledThreadPoolExecutor.java:326)
    at java.util.concurrent.ScheduledThreadPoolExecutor.schedule(ScheduledThreadPoolExecutor.java:533)
    at java.util.concurrent.ScheduledThreadPoolExecutor.execute(ScheduledThreadPoolExecutor.java:622)
    at io.grpc.internal.ManagedChannelImpl$RealChannel$PendingCall.reprocess(ManagedChannelImpl.java:1089)
    at io.grpc.internal.ManagedChannelImpl$RealChannel.updateConfigSelector(ManagedChannelImpl.java:1022)
    at io.grpc.internal.ManagedChannelImpl$NameResolverListenerNamesResolved.run(ManagedChannelImpl.java:1729)
    at io.grpc.SynchronizationContext.drain(SynchronizationContext.java:95)
    at io.grpc.SynchronizationContext.execute(SynchronizationContext.java:127)
    at io.grpc.internal.ManagedChannelImpl$NameResolverListener.onResult(ManagedChannelImpl.java:1815)
    at io.grpc.internal.DnsNameResolver$Resolve.run(DnsNameResolver.java:333)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
    at java.lang.Thread.run(Thread.java:748)

示例伪代码

public class Engine {
    public void startAsync() {
      ...
      OperationFuture future = googleClient.doAsyncRequest();
      // calling future.get(); here works fine
      this.operationFuture = future;
      ...
    }
}

public class Main {
  public static void main(String[] args) {
    ...
    Engine eng = new Engine();
    eng.startAsync();
    eng.operationFuture.get(); // this doesn't work
    ...
  }
}

Answer 1

我找到问题的根源了。该操作在堆栈框架内而不在外部起作用的原因是因为创建和管理此 operationFuture 的 googleClient 对象有自己的 ExecutorService 来处理 operationFuture 生命周期。一旦我们从 startAsync() 方法中 return ， googleClient 对象就会超出范围并被释放，这也会释放与其关联的所有线程并破坏 operationFuture 对象。

为了解决这个问题，googleClient 对象必须与 operationFuture 一起在内存中保持活动状态。例如 DocumentAI：

public void startAsync() {
    ...
    this.googleClient = DocumentProcessorServiceClient.create(docAISettings);
    this.operationFuture = this.googleClient.doAsyncRequest();
    return;
}

只要 future 和客户端对象都没有超出范围，现在从外部调用 operationFuture.get() 方法可以正常工作。

我已经尝试为 googleClient 对象提供一个自定义线程池并让它被垃圾收集（即 googleClient 对象死亡但线程运行 operationFuture 没有）但它似乎不起作用，不知道为什么。

在调用堆栈外部调用 OperationFuture 时 SynchronizationContext 出现异常

Exception in SynchronizationContext when calling OperationFuture outside of callstack

java

google-cloud-sdk

cloud-document-ai