从 TPL foreach 循环修改类变量

Question

我有以下控制台应用程序，它将从超过 7 天的存储帐户容器中删除 blob。

我试图通过调用 WriteNumberFilesProcessed() 在控制台屏幕上创建一个反馈循环，一旦 blob 被删除，它就会增加成员变量 _counter，然后写入控制台。

问题是当最后一个任务在 TPL foreach 循环中完成而不是 _counter 成员变量从 29999 到 30000 时，它将是更远的东西，即 29995。

我不确定我需要做什么来确保 TPL foreach 循环获得最新的成员变量？

class Program
{
    static int _counter = 0;
    static int _numBlobsToDelete = 30000;
    static int _blobCount = 0;

    static void Main(string[] args)
    {
        GetOldBlobs();

        Console.ReadLine();
    }

    static async void GetOldBlobs()
    {
        try
        {
            Stopwatch stopWatch = new Stopwatch();

            CloudStorageAccount acc = CloudStorageAccount.Parse("");
            var client = acc.CreateCloudBlobClient();
            var container = client.GetContainerReference("");
            var rollingTotal = 0;

            while (true)
            {
                stopWatch.Restart();
                Console.WriteLine($"{DateTime.Now}:\tGetting blobs older than a week in a batch of {_numBlobsToDelete}...");

                var blobs = container.ListBlobs("", true).OfType<CloudBlockBlob>().Where(b => (DateTime.UtcNow.AddDays(-7) > b.Properties.LastModified.Value.DateTime)).Take(_numBlobsToDelete).ToList();
                _blobCount = blobs.Count();
                _counter = 0;

                Console.WriteLine($"{DateTime.Now}:\tTime taken to get {_numBlobsToDelete} blobs - {stopWatch.Elapsed}");
                Console.WriteLine($"{DateTime.Now}:\tDeleting {_blobCount} blobs...");

                try
                {
                    Parallel.ForEach(blobs, blob => DeleteBlob(blob));
                }
                catch (Exception ex)
                {
                    Console.WriteLine($"{DateTime.Now}:\t{ex.Message}");
                }

                rollingTotal += _blobCount;

                Console.WriteLine($"\n{DateTime.Now}:\tTime taken to delete blobs - {stopWatch.Elapsed}");
                Console.WriteLine($"{DateTime.Now}:\t{rollingTotal} blobs deleted since startup");
                Console.WriteLine($"{DateTime.Now}:\tSleeping for 5 seconds...\n");
                await Task.Delay(5000);
            }
        }
        catch (Exception ex)
        {
            Console.WriteLine($"{DateTime.Now}:\t{ex.Message}");
        }
    }

    public static void DeleteBlob(CloudBlockBlob blob)
    {
        try
        {
            blob.Delete();
            WriteNumberFilesProcessed();
        }
        catch (Exception ex)
        {
            Console.WriteLine($"{DateTime.Now}:\t{ex.Message}\n");
        }
    }

    public static void WriteNumberFilesProcessed()
    {
        _counter++;
        Console.Write($"\r{_counter} of {_blobCount}");
    }
}

Answer 1

您的问题是多个线程同时调用_counter++。这会导致竞争条件。

要修复它，最简单的方法是使用如下Interlocked方法之一：

Interlocked.Increment(ref _counter);

不使用锁定，可能发生的事情是（给定线程 A 和线程 B）：

假设 _counter 的值为 1：

Thread A: reads value of _counter -> 1
Thread B: reads value of _counter -> 1
Thread A: adds 1 to the value it read -> 2
Thread B: adds 1 to the value it read -> 2
Thread A: writes value back to _counter -> 2
Thread B: writes value back to _counter -> 2

现在_counter有了价值2而不是它应该具有的价值，3。