public void Enqueue(CrawlingQueueItem queueItem)
{
var writeQueueProxy = QueueProxiesWrite.FirstOrDefault();
// if too many in queue already and proxy present, add to proxy (global azure queue), else add to local queue
if (writeQueueProxy != null && LocalQueue.Count >= MaxLocalQueueSize)
{
writeQueueProxy.EnqueueAsync(queueItem, QueueCancellationTokenSource) // No need to wait for the operation to complete successfully
.ContinueWith(pred =>
{
if (pred.Status != TaskStatus.RanToCompletion)
{
queueItem.ChangeStatus(CrawlingQueueItem.CrawlingStatuses.InLocalQueue);
lock (LocalQueue)
LocalQueue.Enqueue(queueItem);
}
})
.ContinueWith(task =>
{
if (task.IsFaulted)
{
Trace.TraceError("CrawlingQueue.[CrawlingQueueProxy].EnqueueAsync: Exception while trying to enqueue {0} [{1}]", queueItem.ResourceLink, task.Exception);
}
});
}
else
{
queueItem.ChangeStatus(CrawlingQueueItem.CrawlingStatuses.InLocalQueue);
lock (LocalQueue)
LocalQueue.Enqueue(queueItem);
}
}
public async Task EnqueueAsync(CrawlingQueueItem crawlingQueueItem, CancellationTokenSource cts)
{
// 1. Asign an AsyncState object when creating a task, so that if remote action fails, we still have the object to insert into local queue
// 2. Use a timeout value for remote operation
crawlingQueueItem.ChangeStatus(CrawlingQueueItem.CrawlingStatuses.InProxyQueue);
await _azureCrawlingQueue.AddMessageAsync(
CloudQueueMessage.CreateCloudQueueMessageFromByteArray(crawlingQueueItem.ToByteArray()), //new CloudQueueMessage(crawlingQueueItem.ToByteArray()),
null,
null,
new QueueRequestOptions
{
},
new OperationContext
{
},
cts.Token
);
}
public override Task <bool> OnBeforeEnqueueAsync(CrawlingQueueItem crawlingQueueItem)
{
return(ExecuteOnAllInterceptorsAsync(eventInterceptor =>
eventInterceptor.OnBeforeEnqueueAsync(crawlingQueueItem)
));
}
public async Task ProcessCrawlingQueueAsync(CrawlingQueue crawlingQueue)
{
_crawlingParameters.CancellationTokenSource.Token.Register(() =>
crawlingQueue.QueueCancellationTokenSource.Cancel()
);
var tasksLock = new System.Threading.ReaderWriterLockSlim();
var tasks = new HashSet <Task>();
var queueItemsProcessingSemaphore = new SemaphoreSlim(crawlingQueue.CrawlingConfiguration.MaxSimmultaneousQueueItemsProcessed / 2, crawlingQueue.CrawlingConfiguration.MaxSimmultaneousQueueItemsProcessed);
while (await queueItemsProcessingSemaphore.WaitAsync(crawlingQueue.CrawlingConfiguration.MaxTimeToProcessOneQueueItem))
{
if (crawlingQueue.QueueCancellationTokenSource.IsCancellationRequested)
{
await Task.WhenAll(tasks.ToArray());
// TODO: Move remaining items from local queue to the distributed queue
// TODO: Figure out how to filter out duplicates from the queue? Or should we?
// We will probably have to resort to known urls-based duplicates check
// Because otherwise we will drown in failing sql queries on multiplie machines
Trace.TraceWarning("ProcessCrawlingQueueAsync: Queue cancellation requested. Preventing dequeing of new elements. Processing will be shut down after currently executing items are complete.");
break;
}
var queueItem = await crawlingQueue.DequeueAsync();
if (queueItem == null) // Both Local and Proxy queues are depleted
{
// NOTE: If Queue is depleted, we must wait until all running tasks are executed, because they might add new items to queue
await Task.WhenAll(tasks.ToArray());
// wait for all queue proxies to complete fetching items
// TODO: consider locking (multithreading scenario)
var queueProxiesPending = crawlingQueue.QueueProxies.Where(queueProxy => queueProxy.IsPending()).ToArray();
if (queueProxiesPending.Length > 0)
{
continue;
}
if (crawlingQueue.LocalQueue.Count > 0)
{
continue;
}
break;
}
if (!await _crawlingEventInterceptorManager.OnAfterDequeueAsync(queueItem))
{
// If interceptor returns false, means it's an instruction to ignore this item;
continue;
}
tasksLock.EnterWriteLock();
queueItem.ChangeStatus(CrawlingQueueItem.CrawlingStatuses.Downloading);
tasks.Add(System.Threading.Tasks.TaskExtensions.Unwrap(
CrawlAsync(queueItem.ResourceLink)
.ContinueWith(async task =>
{
tasksLock.EnterWriteLock();
tasks.Remove(task); // to avoid infinite bloating of the collection
tasksLock.ExitWriteLock();
try
{
queueItem.ChangeStatus(CrawlingQueueItem.CrawlingStatuses.Downloaded);
if (task.Status == TaskStatus.RanToCompletion)
{
var resourceContentUnits = task.Result;
var httpResultUnit = resourceContentUnits.OfType <HttpResultUnit>().Single();
queueItem.ChangeStatus(CrawlingQueueItem.CrawlingStatuses.Processing);
var resourceContentUnitsProcessingCountdown = new AsyncCountdownEvent(resourceContentUnits.Count);
// Process resource content units extracted from Response
foreach (var resourceContentUnit in resourceContentUnits)
{
switch (resourceContentUnit)
{
case ExtractedLinksUnit extractedLinksUnit:
if (extractedLinksUnit.ExtractedLinks.Count > 0)
{
var linksProcessingCountdown = new AsyncCountdownEvent(extractedLinksUnit.ExtractedLinks.Count);
foreach (var extractedLink in extractedLinksUnit.ExtractedLinks)
{
var crawlingQueueItem = new CrawlingQueueItem(extractedLink);
// Do not enqueue item if prevented by any interceptor
if (!await _crawlingEventInterceptorManager.OnBeforeEnqueueAsync(crawlingQueueItem))
{
continue;
}
crawlingQueueItem.ProcessingCompleted += () =>
linksProcessingCountdown.AddCount(1)
;
crawlingQueue.Enqueue(crawlingQueueItem);
}
// Wait while all links are processed before releasing the content units semaphore and set Status = Processed for parent
linksProcessingCountdown.WaitAsync()
.ContinueWith(linksProcessingTask =>
resourceContentUnitsProcessingCountdown.AddCount(1)
);
}
else
{
resourceContentUnitsProcessingCountdown.AddCount(1);
}
// Set Processed status when all extracted links are processed
break;
case ExtractedDataUnit extractedDataUnit:
if (!await _crawlingEventInterceptorManager.OnDataDocumentDownloadedAsync(
queueItem.ResourceLink, // May be a DocumentLink, or a FrameLink. Not quite intuitive and probably requires redesign.
extractedDataUnit,
httpResultUnit
))
{
// If any of interceptors failed to process the download result,
// AND failed to store download result for later processing
// we must re-enqueue the item, in order to ensure the results are not lost for good
// We ignore the item and log the error. Chances are we couldn't process the item for a reason. And repeating would just make it stuck infinitely (re-downloading and re-processing)
// (WAS) we must re-enqueue the item, in order to ensure the results are not lost for good
//crawlingQueue.EnqueueAsync(queueItem);
}
resourceContentUnitsProcessingCountdown.Signal();
break;
case DownloadedFilesUnit downloadedFileUnit:
// If download file is a result of redirection,
// we must either explicitly declare that we're expecting a file, or throw a processing exception
var fileLink = queueItem.ResourceLink as FileLink;
if (fileLink == null)
{
Trace.TraceError($"ProcessCrawlingQueueAsync: Downloaded file unit. Resource link is of type {queueItem.ResourceLink.GetType().Name}, expecting FileLink. Preventing processing.");
break;
}
if (!await _crawlingEventInterceptorManager.OnFileDownloadedAsync(
fileLink,
downloadedFileUnit,
httpResultUnit
))
{
// If any of interceptors failed to process the download result,
// AND failed to store download result for later processing....
// We ignore the item and log the error. Chances are we couldn't process the item for a reason. And repeating would just make it stuck infinitely (re-downloading and re-processing)
// (WAS) we must re-enqueue the item, in order to ensure the results are not lost for good
//crawlingQueue.EnqueueAsync(queueItem);
}
resourceContentUnitsProcessingCountdown.Signal();
break;
case HttpResultUnit httpResultUnitStub:
// TODO: Determine what we should do if HTTP download failed. Either re-enqueue or ignore, or alert/do something else
switch (httpResultUnitStub.HttpStatus)
{
//case HttpStatusCode.InternalServerError: // it's likely to repeat within the same run
case HttpStatusCode.GatewayTimeout:
case HttpStatusCode.RequestTimeout:
queueItem.ChangeStatus(CrawlingQueueItem.CrawlingStatuses.NotLinked);
crawlingQueue.Enqueue(queueItem); // Trying to recrawl item if it failed for some intermitent reason
break;
default:
// We need to invoke ProcessingCompleted only after Data and Links extracted are really processed.
//queueItem.ChangeStatus(CrawlingQueueItem.CrawlingStatuses.ProcessingCompleted);
break;
}
resourceContentUnitsProcessingCountdown.Signal();
break;
default:
throw new NotSupportedException();
}
}
// Do not actually wait for related resources processing completion.
// Those might be extracted links or files. No need to hold queue resources while linked units are downloaded
// Set Processed status after all content units were registered and interceptors triggered
await resourceContentUnitsProcessingCountdown.WaitAsync()
.ContinueWith(resourceContentUnitsProcessingTask =>
queueItem.ChangeStatus(CrawlingQueueItem.CrawlingStatuses.Processed)
);
}
else
{
Trace.TraceError("CrawlAsync: Failed for queue item {0} with exception [{1}]", queueItem.ResourceLink, task.Exception);
}
}
finally
{
queueItemsProcessingSemaphore.Release();
}
})
)
);
tasksLock.ExitWriteLock();
}
await Task.WhenAll(tasks.ToArray());
}
public override async Task <bool> OnBeforeEnqueueAsync(CrawlingQueueItem crawlingQueueItem)
{
return(AddKnownData(crawlingQueueItem.ResourceLink.Uri.ToString()));
}
/// <returns>True - continue item execution, False - ignore this item</returns>
public async virtual Task <bool> OnAfterDequeueAsync(CrawlingQueueItem crawlingQueueItem)
{
return(true);
}
/// <returns>True - continue enqueue, False - prevent enqueue</returns>
public async virtual Task <bool> OnBeforeEnqueueAsync(CrawlingQueueItem crawlingQueueItem)
{
return(true);
}