Great question! Let's break this down step by step:

Can you push the message back to the SNS Topic with a 5-minute delay? #

Short answer: No, you can't do this directly with SNS. Amazon SNS doesn't support delayed message delivery—any message you publish to a topic will be delivered to subscribers immediately. So pushing the message back would just trigger your Lambda right away, which doesn't solve your 5-minute retry requirement.

No-code/low-code AWS patterns to handle this (still using your existing Lambda for data transfer) #

Here are a few straightforward options that don't require heavy additional coding:

1. Use Amazon SQS Delay Queues

This is probably the simplest approach if you want minimal changes:

• Instead of having your Lambda respond directly to SNS, first set up an SQS queue as a subscriber to your SNS topic. (Alternatively, have your initial Lambda send the message to SQS if you still need the SNS -> Lambda trigger for initial checks.)
• When your Lambda checks the third-party API and finds validation isn't complete, use the AWS SDK to send the same message to an SQS queue configured with a 5-minute delay (set via the DelaySeconds parameter when sending the message, or set a default delay on the queue itself—max delay is 15 minutes, which fits your needs).
• Configure this delayed SQS queue to trigger your same Lambda function.
• Each time the Lambda receives the message from SQS, it re-checks the API. Once validation is done, proceed with the S3 transfer; if not, send it back to the delayed queue again (make sure to set a maximum retry limit to avoid infinite loops, like 6 retries for a total of 30 minutes).

2. Use AWS Step Functions for Orchestration

Step Functions is perfect for workflow scenarios that require waiting, retries, and conditional logic—and you can set this up with minimal code (even using the visual designer in the AWS Console):

• Create a Step Functions state machine that starts when it receives the SNS message (you can set up SNS to trigger the state machine directly).
• The state machine first invokes your Lambda to check the third-party validation status.
• If validation is complete, it invokes your existing data transfer Lambda.
• If not, the state machine enters a Wait state configured for 5 minutes, then loops back to the validation check step.
• You can easily set maximum retry attempts or a timeout (e.g., 30 minutes) in the state machine configuration to prevent infinite loops.
• Best part: Your data transfer Lambda stays exactly as it is—you don't need to modify it at all.

3. Lambda Asynchronous Invocation with Retry Configuration

If you want to stick with Lambda alone, you can leverage its built-in async retry capabilities:

• Configure your Lambda to be invoked asynchronously by SNS (this is the default if you set SNS as a trigger, but double-check the invocation type).
• When your Lambda checks the API and finds validation isn't complete, throw a retryable exception (e.g., a custom exception like ValidationInProgressException).
• In the Lambda console, go to the Configuration > Asynchronous invocation settings, and set up a retry policy:
  • Set the minimum retry interval to 5 minutes.
  • Set the maximum number of retries (e.g., 6 retries for 30 minutes total).
  • Specify that only your custom retryable exception triggers a retry.
• Lambda will automatically re-invoke your function after the 5-minute delay each time until validation completes or the retry limit is hit.

Each of these options avoids writing complex retry logic from scratch, and lets you keep using your existing Lambda for the actual S3 data transfer work.

内容的提问来源于stack exchange，提问作者Punter Vicky