A New Set of APIs for Amazon SQS Lifeless-Letter Queue Redrive

As we speak, we launch a brand new set of APIs for Amazon Easy Queue Service (Amazon SQS). These new APIs assist you to handle dead-letter queue (DLQ) redrive programmatically. Now you can use the AWS SDKs or the AWS Command Line Interface (AWS CLI) to programmatically transfer messages from the DLQ to their authentic queue, or to a customized queue vacation spot, to try to course of them once more. A DLQ is a queue the place Amazon SQS robotically strikes messages that aren’t appropriately processed by your client software.

To totally recognize how this new API may assist you to, let’s have a fast look again at historical past.

Message queues are an integral a part of trendy software architectures. They permit builders to decouple companies by permitting asynchronous and message-based communications between message producers and customers. In most methods, messages are continued in shared storage (the queue) till the patron processes them. Message queues enable builders to construct purposes which are resilient to momentary service failure. They assist prioritize message processing and scale their fleet of employee nodes that course of the messages. Message queues are additionally widespread in event-driven architectures.

Asynchronous message alternate will not be new in software architectures. The idea of exchanging messages asynchronously between purposes appeared within the Sixties and was first made widespread when IBM launched TCAM for OS/360 in 1972. The overall adoption got here 20 years later with IBM MQ Sequence in 1993 (now IBM MQ) and when Solar Microsystems launched Java Messaging Service (JMS) in 1998, an ordinary API for Java purposes to work together with message queues.

AWS launched Amazon SQS on July 12, 2006. Amazon SQS is a extremely scalable, dependable, and elastic queuing service that “simply works.” As Werner wrote on the time: “We’ve got chosen a concurrency mannequin the place the method engaged on a message robotically acquires a leased lock on that message; if the message will not be deleted earlier than the lease expires, it turns into out there for processing once more. Makes failure dealing with quite simple.”

On January 29, 2014, we launched dead-letter queues (DLQ). DLQs assist you to keep away from a message that didn’t be processed from staying ceaselessly on prime of the queue, presumably stopping different messages within the queue from processing. With DLQs, every queue has an related property telling Amazon SQS what number of occasions a message could also be introduced for processing (maxReceiveCount). Every message additionally has an related obtain counter (ReceiveCount). Every time a client software picks up a message for processing, the message obtain depend is incremented by 1. When ReceiveCount > maxReceiveCount, Amazon SQS strikes the message to your designated DLQ for human evaluation and debugging. You usually affiliate alarms with the DLQ to ship notifications when such occasions occur. Typical causes to maneuver a message to the DLQ are as a result of they’re incorrectly formatted, there are bugs within the client software, or it takes too lengthy to course of the message.

At AWS re:Invent 2021, AWS introduced dead-letter queue redrive on the Amazon SQS console. The redrive addresses the second a part of the failed message lifecycle. It means that you can reinject the message in its authentic queue to try processing it once more. After the patron software is mounted and able to devour the failed messages, you may redrive the messages from the DLQ again within the supply queue or a custom-made queue vacation spot. It simply requires a few clicks on the console.

As we speak, we’re including APIs permitting you to jot down purposes and scripts that deal with the redrive programmatically. There isn’t any longer a must have a human clicking on the console. Utilizing the API will increase the scalability of your processes and reduces the chance of human error.

Let’s See It in Motion
To check out this new API, I open a terminal for a command-line solely demo. Earlier than I get began, I make sure that I’ve the newest model of the AWS CLI. On macOS I enter brew improve awscli.

I first create two queues. One is the dead-letter queue, and the opposite is my software queue:

# First, I create the dead-letter queue (discover the -dlq I select so as to add on the finish of the queue title)
➜ ~ aws sqs create-queue 
            --queue-name awsnewsblog-dlq                                            
{
    "QueueUrl": "https://sqs.us-east-2.amazonaws.com/012345678900/awsnewsblog-dlq"
}

# second, I retrieve the Arn of the queue I simply created
➜  ~ aws sqs get-queue-attributes 
             --queue-url https://sqs.us-east-2.amazonaws.com/012345678900/awsnewsblog-dlq 
             --attribute-names QueueArn
{
    "Attributes": {
        "QueueArn": "arn:aws:sqs:us-east-2:012345678900:awsnewsblog-dlq"
    }
}

# Third, I create the applying queue. I enter a redrive coverage: publish messages within the DLQ after three supply makes an attempt
➜  ~ aws sqs create-queue 
             --queue-name awsnewsblog 
             --attributes '{"RedrivePolicy": "{"deadLetterTargetArn":"arn:aws:sqs:us-east-2:012345678900:awsnewsblog-dlq","maxReceiveCount":"3"}"}' 
{
    "QueueUrl": "https://sqs.us-east-2.amazonaws.com/012345678900/awsnewsblog"
}

Now that the 2 queues are prepared, I publish a message to the applying queue:

➜ ~ aws sqs send-message 
            --queue-url https://sqs.us-east-2.amazonaws.com/012345678900/awsnewsblog 
            --message-body "Whats up World"
{
"MD5OfMessageBody": "b10a8db164e0754105b7a99be72e3fe5",
"MessageId": "fdc26778-ce9a-4782-9e33-ae73877cfcb2"
}

Subsequent, I devour the message, however I don’t delete it from the queue. This simulates a crash within the message client software. Message customers are alleged to delete the message after profitable processing. I set the maxReceivedCount property to three after I entered the redrivePolicy. I due to this fact repeat this operation 3 times to drive Amazon SQS to maneuver the message to the dead-letter queue after three supply makes an attempt. The default visibility timeout is 30 seconds, so I’ve to attend 30 seconds or extra between the retries.

➜ ~ aws sqs receive-message 
            --queue-url https://sqs.us-east-2.amazonaws.com/012345678900/awsnewsblog
{
"Messages": [
{
"MessageId": "fdc26778-ce9a-4782-9e33-ae73877cfcb2",
"ReceiptHandle": "AQEBP8yOfgBlnjlkGXjyeLROiY7xg7cZ6Znq8Aoa0d3Ar4uvTLPrHZptNotNfKRK25xm+IU8ebD3kDwZ9lja6JYs/t1kBlwiNO6TBACN5srAb/WggQiAAkYl045Tx3CvsOypbJA3y8U+MyEOQRwIz6G85i7MnR8RgKTlhOzOZOVACXC4W8J9GADaQquFaS1wVeM9VDsOxds1hDZLL0j33PIAkIrG016LOQ4sAntH0DOlEKIWZjvZIQGdlRJS65PJu+I/Ka1UPHGiFt9f8m3SR+Y34/ttRWpQANlXQi5ByA47N8UfcpFXXB5L30cUmoDtKucPewsJNG2zRCteR0bQczMMAmOPujsKq70UGOT8X2gEv2LfhlY7+5n8z3yew8sdBjWhVSegrgj6Yzwoc4kXiMddMg==",
"MD5OfBody": "b10a8db164e0754105b7a99be72e3fe5",
"Body": "Hello World"
}
]
}

# wait 30 seconds,
# then repeat two occasions (for a complete of three receive-message API calls)

After three processing makes an attempt, the message will not be within the queue anymore:

➜  ~ aws sqs receive-message 
             --queue-url  https://sqs.us-east-2.amazonaws.com/012345678900/awsnewsblog
{
    "Messages": []
}

The message has been moved to the dead-letter queue. I examine the DLQ to substantiate (discover the queue URL ending with -dlq):

➜  ~ aws sqs receive-message 
             --queue-url  https://sqs.us-east-2.amazonaws.com/012345678900/awsnewsblog-dlq
{
    "Messages": [
        {
            "MessageId": "fdc26778-ce9a-4782-9e33-ae73877cfcb2",
            "ReceiptHandle": "AQEBCLtBMoZYVMMq7fUGNHeCliqE3mFXnkuJ+nOXLK1++uoXWBG31nDejCpxElmiBZWfbcfGJrEdKj4P9HJdrQMYDbeSqB+u1ZlB7CYzQBiQps4SEG0biEoubwqjQbmDZlPrmkFsnYgLD98D1XYWk/Ik6Z2n/wxDo9ko9rbZ15izK5RFnbwveNy8dfc6ireqVB1EGbeGkHcweHGuoeKWXEab1ynZWhNqZsQgCR6pWRkgtn59lJcLv4cJ4UMewNzvt7tMHH69GvVjXdYDYvJJI2vj+6RHvcvSHWWhTNT+CuPEXguVNuNrSya8gho1fCnKpVwQre6HhMlLPjY4wvn/tXY7+5rmte9eXagCqLQXaENB2R7qWNVPiWRIJy8/cTf37NLYVzBom030DNJlH9EeceRhCQ==",
            "MD5OfBody": "b10a8db164e0754105b7a99be72e3fe5",
            "Body": "Hello World"
        }
    ]
}

Now that the setup is prepared, let’s programmatically redrive the message to its authentic queue. Let’s assume I perceive why the patron didn’t appropriately course of the message and that I mounted the patron software code. I take advantage of start-message-move-task on the DLQ to start out the asynchronous redrive. There’s an non-compulsory attribute (MaxNumberOfMessagesPerSecond) to manage the speed of the redrive:

➜ ~ aws sqs start-message-move-task 
            --source-arn arn:aws:sqs:us-east-2:012345678900:awsnewsblog-dlq
{
    "TaskHandle": "eyJ0YXNrSWQiOiI4ZGJmNjBiMy00MmUwLTQzYTYtYjg4Zi1iMTZjYWRjY2FkNmEiLCJzb3VyY2VBcm4iOiJhcm46YXdzOnNxczp1cy1lYXN0LTI6NDg2NjUyMDY2NjkzOmF3c25ld3NibG9nLWRscSJ9"
}

I can checklist and examine standing the of the transfer duties I initiated with list-message-move-tasks or cancel a working activity by calling the cancel-message-move-task API:

➜ ~ aws sqs list-message-move-tasks 
            --source-arn arn:aws:sqs:us-east-2:012345678900:awsnewsblog-dlq
{
    "Outcomes": [
        {
            "Status": "COMPLETED",
            "SourceArn": "arn:aws:sqs:us-east-2:012345678900:awsnewsblog-dlq",
            "ApproximateNumberOfMessagesMoved": 1,
            "ApproximateNumberOfMessagesToMove": 1,
            "StartedTimestamp": 1684135792239
        }
    ]
}

Now my software can devour the message once more from the applying queue:

➜  ~ aws sqs receive-message 
             --queue-url  https://sqs.us-east-2.amazonaws.com/012345678900/awsnewsblog                                   
{
    "Messages": [
        {
            "MessageId": "a7ae83ca-cde4-48bf-b822-3d4bc1f4dcae",
            "ReceiptHandle": "AQEB9a+Dm2nvb3VUn9+46j9UsDidU/W6qFwJtXtNWTyfoSDOKT7h73e6ctT9RVZysEw3qqzJOx1cxblTTOSrYwwwoBA2qoJMGsqsrsRGGYojBvf9X8hqi8B8MHn9rTm8diJ2wT2b7WC+TDrx3zIvUeiSEkP+EhqyYOvOs7Q9aETR+Uz02kQxZ/cUJWsN4MMSXBejwW+c5ivv5uQtpfUrfZuCWa9B9O67Kj/q52clriPHpcqCCfJwFBSZkGTXYwTpnjxD4QM7DPS+xVeVfTyM7DsKCAOtpvFBmX5m4UNKT6TROgCnGxTRglUSMWQp8ufVxXiaUyM1dwqxYekM9uX/RCb01gEyCZHas4jeNRV5nUJlhBkkqPlw3i6w9Uuc2y9nH0Df8nH3g7KTXo4lv5Bl3ayh9w==",
            "MD5OfBody": "b10a8db164e0754105b7a99be72e3fe5",
            "Body": "Hello World"
        }
    ]
}

Availability
DLQ redrive APIs can be found as we speak in all business Areas the place Amazon SQS is out there.

Redriving the messages from the dead-letter queue to the supply queue or a customized vacation spot queue generates further API calls billed primarily based on current pricing (beginning at $0.40 per million API calls, after the primary million, which is free each month). Amazon SQS batches the messages whereas redriving them from one queue to a different. This makes shifting messages from one queue to a different a easy and low-cost possibility.

To study extra about DLQ and DLQ redrive, examine our documentation.

Do not forget that we reside in an asynchronous world—so ought to your purposes. Get began as we speak and write your first redrive software.

— seb