Hey everyone,
I'm dealing with a race condition issue when pushing user information to a SaaS product daily. We're using AWS Lambda with a concurrency setting of 10, but it seems like the SaaS API is having trouble processing these requests correctly due to the concurrent calls. Has anyone faced a similar situation, and what solutions have you found effective?
3 Answers
Have you considered using a pub/sub pattern or queues? They can help manage the flow of requests and reduce the chances of racing issues.
Depending on why you're experiencing these race conditions, there are several strategies you could try. For example, if you're pushing updates to the same customer in a SaaS CRM, you might run into conflicts with multiple requests. One solution could be implementing a FIFO (First In, First Out) queue using message group IDs based on the customer ID. This way, you can handle multiple customers at once, but ensure that requests for the same customer are processed in order. However, if that doesn't work for you, you might have to throttle the requests down to a concurrency of 1, which can prevent racing but might lead to a backlog if the queue builds up too much.
If preserving the order of updates is crucial for your application, dropping the concurrency to 1 might be the way to go. Alternatively, you could set up an SQS FIFO queue in front of your Lambda function, using message group IDs based on user IDs. This would allow for ordering per user without decreasing the processing speed for requests from different users. The race condition is happening because you're sending multiple updates for the same entity at the same time.
Exactly! A FIFO queue with message group IDs for each customer would help maintain order while allowing concurrency for different customers. Using SQS FIFO with Lambda is a solid approach.