繁体   English   中英

C# 处理 IAsyncEnumerable 项目与 asp.net 核心托管服务消费者中的多个并行任务

[英]C# process IAsyncEnumerable items with several parallel tasks in asp.net core hosted service consumer

我正在使用托管服务在 asp.net 核心应用程序中实现生产者-消费者。 我能够让它工作到消费者同步处理来自_recordProcessingChannel.ReadAllAsync()的项目的地步。

我正在尝试将_recordProcessingChannel.ReadAllAsync()的结果拆分为几个并行任务。 例如:我从频道读取了 10000 个项目,我想将这项工作分成 4 个单独的任务,每个 ICMService 处理 2500 个项目。

消费者:

await foreach (var record in _recordProcessingChannel.ReadAllAsync())
{

    using var scope = _serviceProvider.CreateScope();
    var processor = scope.ServiceProvider.GetRequiredService<ICMService>();

    processor.UploadRecord(record);
                
}

读者:

public IAsyncEnumerable<RecordData> ReadAllAsync(CancellationToken ct = default) => _channel.Reader.ReadAllAsync(ct);

预先感谢您提供的任何帮助

您可以启动所需数量的处理任务并使用BlockingCollection将工作enqueue 像这样的东西:

// my dummy async enumerable
public async IAsyncEnumerable<int> ReadAllAsync()
{
    for (int i = 0; i < 3; i++)
    {
        yield return i*3 + 1;
        yield return i*3 + 2;
        yield return i*3 + 3;
        await Task.Delay(200);
    }
    yield return 777;
}
var collection = new BlockingCollection<int>();
// start "processors"
var tasks = Enumerable.Range(0, 4)
    .Select(i => 
        Task.Run(() =>
        {
            while (!collection.IsCompleted)
            {           
                int? data = null;
                try
                {
                    data = collection.Take();
                }
                catch (InvalidOperationException) { }

                if (data != null)
                {
                    // simulate processing 
                    Thread.Sleep(400);
                    Console.WriteLine(data.Value);
                }
            }
            Console.WriteLine("No more items to take.");
        }))
    .ToArray();

await foreach (var record in ReadAllAsync())
{
    collection.Add(record);
}
collection.CompleteAdding(); // signal that enqueuing has finished

await Task.WhenAll(tasks);

这可以通过引入一些异步信号(例如SemaphoreSlim.WaitAsyncAsyncManualResetEvent.WaitAsync )来改进,因此消费者线程在等待新项目时不会消耗 CPU。 例如:

var collection = new ConcurrentQueue<int>();
var semaphore = new SemaphoreSlim(0, 4);
var cts = new CancellationTokenSource(); // to signal that queueing is completed
var tasks = Enumerable.Range(0, 4)
    .Select(i => 
        Task.Run(async () =>
        {
            while (true)
            {
                if (cts.Token.IsCancellationRequested && !collection.Any())
                {
                    Console.WriteLine("No more items to take.");
                    break;
                }
                else if (!cts.Token.IsCancellationRequested)
                {
                    try
                    {
                        await semaphore.WaitAsync(cts.Token);
                    }
                    catch (OperationCanceledException)
                    {
                        //ignore
                    }
                }

                if(collection.TryDequeue(out var data))
                {
                    //simulate work
                    Thread.Sleep(400);
                    Console.WriteLine(data);
                }                   
            }
        }))
    .ToArray();

await foreach (var record in ReadAllAsync())
{
    collection.Enqueue(record);
    semaphore.Release();
}
cts.Cancel(); // addition completed.
await Task.WhenAll(tasks);
Console.WriteLine("end");

暂无
暂无

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM