Recursive loops Detected and Stopped for Lambda Function guidance in our scenario

Question

We have deployed a standard AWS SQS queue.This queue is configured to retain messages for 4 days.

**Message processing:** Once message is received by the consumer(lambda with reserve concurrency = 5), consumer checks where the message processable or not. If not,we update the message visibility time and terminate the process without deleting the message by throwing error(assuming that the termination will retain the message).

**Doubts:**

1. Retention period is set to 4 days but sometimes message stays in queue for longer time. How is it possible?

2. How ApproximateReceiveCount is maintained. We observed that there are gaps in values in sequential message processing. for eg. ApproximateReceiveCount = 9 in one message and same value is set to 3 in previous message for same messageID.

3. AWS documentation says recursive loop detection happens after 16th iteration, however we have observed even with ApproximateReceiveCount = 18, we have not got any alert.

Answer

How did you check that retention is not accurate?

Due to the distributed nature of SQS, ApproximateReceiveCount is only an approximation (as it names implies). Different hosts can have different values.

What you are doing is not consider recursive invocation. A recursive use case if if your function sends the messages back to the same queue. This can cause an infinite loop. Not deleting the message from the queue is not a recursive operation as eventually it will either hit the maximum retries or will expire from the queue.

Instead of failing your function, you can just return an array with [partial batch responses](https://docs.aws.amazon.com/lambda/latest/dg/with-sqs.html#services-sqs-batchfailurereporting).

Recursive loops Detected and Stopped for Lambda Function guidance in our scenario

Relevant content