Event-driven workers

When a request shouldn’t block on slow work, you publish an event and let workers drain it. This is the fan-out pattern: an API Lambda publishes to SNS, which fans into SQS queues, each drained by a worker Lambda.

Build it

1. Wire Internet → API Gateway → Lambda (the producer) — see the serverless API guide for that base.
2. Add an SNS topic and wire the producer Lambda’s downstream socket to it.
3. Add an SQS queue and subscribe it to the SNS topic (the subscription is generated for you).
4. Add a worker Lambda and wire the SQS queue to its event-source socket — this emits an aws_lambda_event_source_mapping, so the worker is invoked per batch.
5. Optionally add an EventBridge rule for scheduled (cron) invocations.

Why the sockets matter

The SQS → worker connection generates the event source mapping automatically — you never hand-write the polling wiring. And because the queue is wired to a specific worker, adding a second queue and a second worker stays unambiguous.

resource "aws_sns_topic" "events" { name = "events" }

resource "aws_sqs_queue" "work" { name = "work" }

resource "aws_sns_topic_subscription" "fanout" {
  topic_arn = aws_sns_topic.events.arn
  protocol  = "sqs"
  endpoint  = aws_sqs_queue.work.arn
}

resource "aws_lambda_event_source_mapping" "drain" {
  event_source_arn = aws_sqs_queue.work.arn
  function_name    = aws_lambda_function.worker.arn
  batch_size       = 10
}

Tip

Add a dead-letter queue to the worker so failed messages are retained instead of lost — wire a second SQS queue to the worker’s dead-letter socket.

Next

• Simulate the fan-out under load to size the worker concurrency.
• See the messaging nodes in the node reference.