我在多线程环境中大量抓取网页的内容。我需要一个可靠的下载器组件，它可以容忍临时服务器故障、连接断开等。下面是我的代码的样子。
现在，我一遍又一遍地遇到一个奇怪的情况：一切都开始得很完美。10 个线程同时拉取数据大约 10 分钟。在那之后，我在调用我的请求对象的 GetResponse 方法后立即开始收到带有超时的 WebException 。休息一下（让线程入睡）没有帮助。只有当我停止应用程序并重新启动它直到接下来的 10 分钟过去并且问题再次出现时，它才会有所帮助。
我已经尝试过但没有任何帮助：
- 通过“使用”语句显式关闭/处理响应对象
- 调用 request.Abort 它可以帮助的任何地方
- 在 ServicePointManager/ServicePoint 和 WebRequest 级别操作超时（延长/缩短超时间隔）
- 操作 KeepAlive 属性
- 调用 CloseConnectionGroup
- 操纵同时运行的线程数
没有什么帮助！因此，这似乎是一个错误，或者至少是记录非常差的行为。我在 Google 和 Stackoverflow 上看到了很多关于此的问题，但没有一个得到完全回答。基本上人们会从上面的列表中提出其中一项建议。我都试过了。
public TResource DownloadResource(Uri uri)
{
for (var resourceReadingAttempt = 0; resourceReadingAttempt <= MaxTries; resourceReadingAttempt++)
{
var request = (HttpWebRequest)WebRequest.Create(uri);
HttpWebResponse response = null;
for (var downloadAttempt = 0; downloadAttempt <= MaxTries; downloadAttempt++)
{
if (downloadAttempt > 0)
{
var sleepFor = TimeSpan.FromSeconds(4 << downloadAttempt) + TimeSpan.FromMilliseconds(new Random(DateTime.Now.Millisecond).Next(1000));
Trace.WriteLine("Retry #" + downloadAttempt + " in " + sleepFor + ".");
Thread.Sleep(sleepFor);
}
Trace.WriteLine("Trying to get a resource by URL: " + uri);
var watch = Stopwatch.StartNew();
try
{
response = (HttpWebResponse)request.GetResponse();
break;
}
catch (WebException exception)
{
request.Abort();
Trace.WriteLine("Failed to get a resource by the URL: " + uri + " after " + watch.Elapsed + ". " + exception.Message);
if (exception.Status == WebExceptionStatus.Timeout)
{
//Trace.WriteLine("Closing " + request.ServicePoint.CurrentConnections + " current connections.");
//request.ServicePoint.CloseConnectionGroup(request.ConnectionGroupName);
//request.Abort();
continue;
}
else
{
using (var failure = exception.Response as HttpWebResponse)
{
Int32 code;
try { code = failure != null ? (Int32)failure.StatusCode : 500; }
catch { code = 500; }
if (code >= 500 && code < 600)
{
if (failure != null) failure.Close();
continue;
}
else
{
Trace.TraceError(exception.ToString());
throw;
}
}
}
}
}
if (response == null) throw new ApplicationException("Unable to get a resource from URL \"" + uri + "\".");
try
{
// response disposal is required to eliminate problems with timeouts
// more about the problem: http://stackoverflow.com/questions/5827030/httpwebrequest-times-out-on-second-call
// http://social.msdn.microsoft.com/Forums/en/netfxnetcom/thread/a2014f3d-122b-4cd6-a886-d619d7e3140e
TResource resource;
using (var stream = response.GetResponseStream())
{
try
{
resource = this.reader.ReadFromStream(stream);
}
catch (IOException exception)
{
Trace.TraceError("Unable to read the resource stream: " + exception.ToString());
continue;
}
}
return resource;
}
finally
{
// recycle as much as you can
if (response != null)
{
response.Close();
(response as IDisposable).Dispose();
response = null;
}
if (request != null)
{
//Trace.WriteLine("closing connection group: " + request.ConnectionGroupName);
//request.ServicePoint.CloseConnectionGroup(request.ConnectionGroupName);
request.Abort();
request = null;
}
}
}
throw new ApplicationException("Resource was not able to be acquired after several attempts.");
}