I have a WCF on a Web Role and then a Worker Role to process the messages added to an azure queue by the WCF.
I am doing the following :
var queue = queueStorage.GetQueueReference("myqueue");
var message = new CloudQueueMessage(string.Format("{0},{1}", pWord,processed));
queue.AddMessage(message);
Then I want to wait until the message has been processed, but I dont know if my queue object will get updated on its own or I have to do something about it.
On my worker role I have the following :
This is my onStart method :
CloudQueueClient queueClient = storageAccount.CreateCloudQueueClient();
inputQueue = queueClient.GetQueueReference("myqueue");
And then on my Run method :
while (true)
{
try
{
// Retrieve and process a new message from the queue.
msg = inputQueue.GetMessage();
if (msg != null)
{
result = processMessage(msg);
On my processMessage method :
var messageParts = msg.AsString.Split(new char[] { ',' });
var word = messageParts[0];
var processed = Convert.ToBoolean(messageParts[2]);
word = "recibido";
processed = true;
addMessageToQueue2(userId,processed);
return 1;
Add message to queue is :
var queue = outputQueue.GetQueueReference("myQueue");
var message = new CloudQueueMessage(string.Format("{0},{1}", pWord, pProcessed));
queue.AddMessage(message);
Im fairly new to queues but I think this should work so all I need is just waiting until the message has been processed but dont know how it internally works.
Not quite sure what you mean by waiting until the message has been processed. With Azure queues, the operation is very simple:
Place messages on queue (with message TTL)
Read message(s) from queue (with processing-timeout). This processing timeout says "I promise to finish dealing with this message, and then deleting this message, before this timeout is hit.
The queue message stays in the queue but becomes invisible to all other readers during the timeout period.
The message-reader then deletes the message from the queue.
Assuming the code that read the queue message deletes the message before the promised timeout expires, all is good in QueueLand. However: If the processing goes beyond the timeout period, then the message becomes visible again. And... if someone else then reads that message, the original reader loses rights to delete it (they'll get an exception when making an attempt).
So: Long story short: You can process a message for as long as you want, within the stated timeout period, and then delete it. If your code crashes during processing, the message will eventually reappear for someone else to read. If you want to deal with poison messages, just look at the DequeueCount property of the message to see how many times it's been read (and if over a certain threshold, do something special with the message, like tuck it away in a blob or table row for future inspection by the development team).
See this article for all documented queue limits (and a detailed side-by-side comparison with Service Bus Queues).
Related
We have a scenario when we pull message from Azure Service Bus Queue and for some reason if one of the down stream is down than we would like to delay a message and put back to queue. I understand we can do through multiple ways(Set the property ScheduledEnqueueTime or use Schedule API)but either way we will have to create a new message and put back to queue which will lose the delivery count and also can result in an issue where we have duplicate message where
sending the clone and completing the original are not an atomic operation and one of them fails.
https://www.markheath.net/post/defer-processing-azure-service-bus-message
based on the above article only way seems to be have our custom property, Is that the only way still as this article was written in 2016.
Scheduling a new message back does not increase the delivery count. And as you said, sending a message and completing a message are not atomic, it can be atomic with the help of transactions, thereby ensuring that all operations belonging to a given group of operations either succeed or fail jointly.
Here's an example:
ServiceBusClient client = new ServiceBusClient("<connection-string>");
ServiceBusReceiver serviceBusReceiver = client.CreateReceiver("<queue>");
ServiceBusSender serviceBusSender = client.CreateSender("<queue>");
var message = await serviceBusReceiver.ReceiveMessageAsync();
// Your condition to handle the down stream
if (true)
{
using (var ts = new TransactionScope(TransactionScopeAsyncFlowOption.Enabled))
{
await serviceBusReceiver.CompleteMessageAsync(message);
var newMessage = new ServiceBusMessage(message);
newMessage.ScheduledEnqueueTime = new DateTimeOffset(DateTime.UtcNow.AddMinutes(1));
await serviceBusSender.SendMessageAsync(newMessage);
ts.Complete();
}
}
(FYI - I am new ASB)
A couple of questions around Azure Service Bus:
How do you get a message from a Queue but leave it there until its' TTL expires? I would have thought simply not calling CompleteMessageAsync would do just that, but it appears to get removed regardless.
How do get a message from a Queue, but only dequeue (remove) it when received by a specific receiver?
Message.ApplicationProperties["ReceiverId"].ToString() == "123"
// now you can remove it
Thanks
How do you get a message from a Queue but leave it there until its' TTL expires?
You can peek at messages rather than receive them. The problem is that the message will be picked up again and again until the delivery count exceeds the maximum and the message will dead-letter, which you don't want to happen. I would review what you're trying to achieve here as it's a contradictory setup. You want the message to have a TTL in anticipation that it's not picked up, but then you want to probe it until TTL expires continuedly.
How do get a message from a Queue, but only dequeue (remove) it when received by a specific receiver?
My advice is don't use a queue for that. If you target a specific destination, express it with your entity topology. For example: publish a message on a topic and have different subscriptions based on the subscriber identification. That way you can have messages for specific subscribers, where a logical subscriber can be scaled out.
1-Use the PeekMessage:
You can peek at the messages in the queue without removing them from
the queue by calling the PeekMessages method. If you don't pass a
value for the maxMessages parameter, the default is to peek at one
message.
//-------------------------------------------------
// Peek at a message in the queue
//-------------------------------------------------
public void PeekMessage(string queueName)
{
// Get the connection string from app settings
string connectionString = ConfigurationManager.AppSettings["StorageConnectionString"];
// Instantiate a QueueClient which will be used to manipulate the queue
QueueClient queueClient = new QueueClient(connectionString, queueName);
if (queueClient.Exists())
{
// Peek at the next message
PeekedMessage[] peekedMessage = queueClient.PeekMessages();
// Display the message
Console.WriteLine($"Peeked message: '{peekedMessage[0].Body}'");
}
}
https://learn.microsoft.com/en-us/azure/storage/queues/storage-dotnet-how-to-use-queues?tabs=dotnet
2-you can also use PeekMessage, check for the property you want (ReceiverId), and it case it's the right one, just complete the message:
// ServiceBusReceiver
await receiver.CompleteMessageAsync(receivedMessage);
What is the proper method of consuming a message, processing it, and then publishing it? I run into a lot of unacknowledged messages and I believe there is some blocking going on. Trying to understand the best practice for something like this.
I'm working on a set of services that will process around 50k requests a day. I have decided to use RabbitMQ and three Windows Services written in Dotnet Core 3.1.
I have diagrammed the process but essentially it works like this:
an external service publishes the message to Queue #1
service A is "listening" on Queue #1 and consumes any messages that arrive in the Queue. A database call is made and then Service A passes message to Queue #2
service B is "listening" on Queue #2 and consumes any messages that arrive in the Queue. Some internal processing is done and then Service B passes message to Queue #3
service C is "listening" on Queue #2 and consumes any messages that arrive in the Queue. Some internal processing is done and then Service C pushes message to database
Code example is below image
protected override void OnStart(string[] args)
{
logger.LogInformation("Starting Service ...");
base.OnStart(args);
string queue = "Queue_StageOne";
this.connection = factory.CreateConnection();
this.channel = connection.CreateModel();
this.publishingChannel = connection.CreateModel();
this.channel.BasicQos(0, 1, false);
consumer = new AsyncEventingBasicConsumer(channel);
consumer.Received += Consumer_Recieved;
this.channel.BasicConsume(queue: queue, autoAck: false, consumer: consumer);
}
private async Task Consumer_Recieved(object sender, BasicDeliverEventArgs #event)
{
var body = #event.Body;
var message = Encoding.UTF8.GetString(body.ToArray());
var inboundTransferObject = PatientObject.ConvertFromJson(message);
//logger.LogInformation("Processed message " + inboundTransferObject.WebhookMessageId);
//ServicePointManager.SecurityProtocol = SecurityProtocolType.SystemDefault;
//X509Certificate2 cert = new X509Certificate2(config["CertificationPath"].ToString(), config["PFXPassword"]);
//JToken access_token = GetAccessToken(cert);
//JObject payerData = GetPractitionerData(inboundTransferObject, cert, access_token);
//inboundTransferObject = ProcessPractitioner(inboundTransferObject, payerData);
var outboundTransferObject = Encoding.ASCII.GetBytes(inboundTransferObject.ConvertToJson());
channel.BasicAck(deliveryTag: #event.DeliveryTag, multiple: false);
publishingChannel.BasicPublish(exchange: "ExchangeA", routingKey: "Queue_StageTwo", basicProperties: null, body:outboundTransferObject);
await Task.Delay(250);
}
It's not clear exactly what you're asking here but one thing that does stand out is that your services should not acknowledge the inbound message unless and until they've completed all their processing steps, and that includes publishing follow-on outbound messages. In your code sample you appear to acknowledge the inbound message before publishing the outbound message.
That however does not explain the symptom you described "I run into a lot of unacknowledged messages". When do you run into these? How many is a lot? Have you set a prefetch limit on your channel? For testing purposes, you could try setting your prefetch count to one to ensure that only one message is in-flight at a time.
channel.BasicQos(1, global: true)
Please see this section of the RabbitMQ documentation:
"Because messages are sent (pushed) to clients asynchronously, there is usually more than one message "in flight" on a channel at any given moment. In addition, manual acknowledgements from clients are also inherently asynchronous in nature. So there's a sliding window of delivery tags that are unacknowledged. Developers would often prefer to cap the size of this window to avoid the unbounded buffer problem on the consumer end. This is done by setting a "prefetch count" value using the basic.qos method. The value defines the max number of unacknowledged deliveries that are permitted on a channel. Once the number reaches the configured count, RabbitMQ will stop delivering more messages on the channel unless at least one of the outstanding ones is acknowledged."
I've been using Azure Storage Queues to post messages too, then write the messages to a db table. However I've noticed that when an error occurs processing messages on the queue, the message is written to a poison queue.
Here is some background to the setup of my app:
Azure Web App -> Writes message to the queue
Azure function -> Queue trigger processes the message and writes the contents to a db
There was an issue with the db schema which caused the INSERTS to fail. Each message was retried 5 times, which I believe is the default for retrying queue messages, and after the 5th attempt the message was placed on the poison queue.
The db schema was subsequently fixed but now I've no way of processing the messages on the poison queue.
My question is can we recover messages written to the poison queue in order to process them and INSERT them into the db, and if so how?
For your particular problem, I would recommend solution mentioned in question part of this post: Azure: How to move messages from poison queue to back to main queue?
Please note that name of poison queue == $"{queueName}-poison"
In my current project I've created something what is called: "Support functions" in the FunctionApp. It exposes a special HTTP endpoint with Admin authorization level that can be executed at any time.
Please See the code below, which solves the problem of reprocessing messages from the poison queue:
public static class QueueOperations
{
[FunctionName("Support_ReprocessPoisonQueueMessages")]
public static async Task<IActionResult> Support_ReprocessPoisonQueueMessages([HttpTrigger(AuthorizationLevel.Admin, "put", Route = "support/reprocessQueueMessages/{queueName}")]HttpRequest req, ILogger log,
[Queue("{queueName}")] CloudQueue queue,
[Queue("{queueName}-poison")] CloudQueue poisonQueue, string queueName)
{
log.LogInformation("Support_ReprocessPoisonQueueMessages function processed a request.");
int.TryParse(req.Query["messageCount"], out var messageCountParameter);
var messageCount = messageCountParameter == 0 ? 10 : messageCountParameter;
var processedMessages = 0;
while (processedMessages < messageCount)
{
var message = await poisonQueue.GetMessageAsync();
if (message == null)
break;
var messageId = message.Id;
var popReceipt = message.PopReceipt;
await queue.AddMessageAsync(message); // a new Id and PopReceipt is assigned
await poisonQueue.DeleteMessageAsync(messageId, popReceipt);
processedMessages++;
}
return new OkObjectResult($"Reprocessed {processedMessages} messages from the {poisonQueue.Name} queue.");
}
}
Alternatively it may be a good idea to create a new message with the additional metadata (as information that the message has already been processed in the past with no success - then it may be send to the dead letter queue).
You have two options
Add another function that is triggered by messages added to the poison queue. You can try adding the contents to the db in this function. More details on this approach can be found here. Of course, if this function too fails to process the message you could check the dequeue count and post a notification that needs manual intervention.
Add an int 'dequeueCount' parameter to the function processing the queue and after say 5 retries log the failure instead of letting the message go the poison queue. For example you can send an email to notify that manual intervention is required.
You can use azure management studio(cerulean) and move the message from poison queue to actual queue. Highly recommended tool to access queues and blobs and do any production related activity also. https://www.cerebrata.com/products/cerulean
I am just user of the tool and no way affiliated, i recommended because it is very powerful, very useful and makes you very productive.
Click on move and message can be moved to the actual uploaded queue
Just point your Azure function to the poison queue and the items in that poison queue will be handled. More details here: https://briancaos.wordpress.com/2018/05/03/azure-functions-how-to-retry-messages-in-the-poison-queue/
Azure Storage Explorer(version above 1.15.0) has now added support to move messages from one queue to another. This makes it possible to move all, or a selected set of messages, from the poison queue back to the original queue.
https://github.com/microsoft/AzureStorageExplorer/issues/1064
I'm trying to resubmit a message from a deadletter queue.
I am can replay a message on a dead letter queue, thats fine.
The problem is when I want to now delete this from the deadletter queue.
Here is what I am trying to do:
var subscription = "mySubscription";
var topic = "myTopic";
var connectionString = "connectionStringOnAzure";
var messagingFactory = MessagingFactory.CreateFromConnectionString(connectionString);
var messageReceiver = messagingFactory.CreateMessageReceiver(SubscriptionClient.FormatDeadLetterPath(topic, subscription), ReceiveMode.ReceiveAndDelete);
long messageSequenceNumber = 835;
var brokeredMessage = messageReceiver.Receive(messageSequenceNumber); // this part fails
// mark message as complete to remove from the queue
brokeredMessage.Complete();
I get following error message:
Microsoft.ServiceBus.Messaging.MessageNotFoundException : Failed to lock one or more specified messages. The message does not exist..TrackingId:ae15edcc-06ac-4d2b-9059-009599cf5c4e_G5_B15,TimeStamp:8/13/2013 1:45:42 PM
However, instead of specifying a message sequence number and I just use the ReceiveBatch as shown below, it is fine.
// this works and does not throw any errors
var brokeredMessages = messageReceiver.ReceiveBatch(10);
Am I missing something? Or is there another way of reprocessing deadletters and removing them?
The deadletter queue is processed in sequence just like any other queue.
The Receive(seqNo) method is used in combination with Defer(), which puts the message into a different secondary Queue - the "deferral queue". The deferral queue exists for scenarios where you are getting messages out of the expected order (eg. in a state machine) and need a place to put the messages that arrived early. Those you can park with Defer() and make a note of that (probably even in session state) and then pull the messages once you're ready to do so. The Workflow Manager runtime used by SharePoint uses that feature, for instance.
After creating receiver you can politely start receiving all messages (without being picky) till you encounter message with your SequenceNumber, call Complete() on the message and stop iterating the queue. i.e
while (true)
{
BrokeredMessage message = receiver.Receive();
if (message.SequenceNumber == sequenceNumber)
{
message.Complete();
break;
}
}
Without completing message it remains in the queue and that's what you want (at least in .NET 4.5. Worth to note that if your Sequence Number is not found Receiver will loop the queue indefinitely.